MAGNETIC

RESONANCE
IMAGING
Physical and Biological Principles
This page intentionally left blank
 FOURTH EDITION

MAGNETIC
RESONANCE
IMAGING
Physical and Biological Principles

Stewart Carlyle Bushong, ScD, FAAPM, FACR

Professor of Radiologic Science
Baylor College of Medicine
Houston, Texas

Geoffrey Clarke, PhD

Professor and Chief of Graduate Education
Department of Radiology
The University of Texas Health Science Center at San Antonio
San Antonio, Texas
3251 Riverport Lane
St. Louis, Missouri 63043

MAGNETIC RESONANCE IMAGING: PHYSICAL AND ISBN: 978-0-323-07354-7

BIOLOGICAL PRINCIPLES
Copyright © 2015, 2003, 1996, 1988 by Mosby, Inc., an affiliate of Elsevier Inc.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic
or mechanical, including photocopying, recording, or any information storage and retrieval system,
without permission in writing from the publisher. Details on how to seek permission, further
information about the Publisher’s permissions policies and our arrangements with organizations
such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our
website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described herein. In
using such information or methods they should be mindful of their own safety and the safety of
others, including parties for whom they have a professional responsibility.
With respect to any drug or pharmaceutical products identified, readers are advised to check
the most current information provided (i) on procedures featured or (ii) by the manufacturer of
each product to be administered, to verify the recommended dose or formula, the method and
duration of administration, and contraindications. It is the responsibility of practitioners, relying
on their own experience and knowledge of their patients, to make diagnoses, to determine dosages
and the best treatment for each individual patient, and to take all appropriate safety precautions.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

Bushong, Stewart C., author.
  Magnetic resonance imaging : physical and biological principles / Stewart Carlyle Bushong, Geoffrey
Clarke.—Fourth edition.
   p. ; cm.
  Includes bibliographical references and index.
  ISBN 978-0-323-07354-7 (pbk. : alk. paper)
  I.  Clarke, Geoffrey David, 1956- author.  II.  Title.
  [DNLM:  1.  Magnetic Resonance Imaging.  2.  Magnetic Resonance Spectroscopy.  WN 185]
  RC78.7.N83
  616.07′54–dc23
2014016502
Executive Content Strategist: Sonya Seigafuse
Content Development Specialist: Kristen Mandava
Publishing Services Manager: Catherine Jackson
Senior Project Manager: Rachel E. McMullen
Design Direction: Maggie Reid
Printed in the United States of America
Last digit is the print number:  9  8  7  6  5  4  3  2  1 
This book is also dedicated to the 429 Baylor College of Medicine radiology resident TOOLS who
endured my medical physics classes, my easy exams, and my lousy jokes since I began doing this in
1967. They gave me much grief but much more pleasure, and most have become lasting close friends.

Asron J. Baxtor ’12—Achal Sarna (HMP) ’04—Achilles Chatziioannou ’96—Adam W. Meyers ’13 –Ahmad A. Khan ’14—Akom Vutpakdi
’71—Alfred E. Delumpa (HMP) ’14—Alicia Roman ’10—Amaia Rauch (HMP) ’08—Amarinthia Curtis (NP) ’08—Aminata Traore ’15—
Anand Prabhakar (NP, HMP) ’09—Andrew J. Marsala ’16—Andrew R. Palisch ’12—Angel Gunn ’05—Anjali Agrawal (HMP) ’02—Ann
Marie Marciel ’08—Anthony Sparks ’14—Anuja Jhingran^ ’93—Aparna Annan ’12—Arnold Nitishian• ’74—Artie Schwartz ’95—Arturo
Castro^ ’82—Ashley Roark ’15—Augusto Merello• ’82, ’85—B. Michael Driver^ ’85—B.A. King ’72—Barry M. Uhl^ ’00—Becky
Borders ’09—Beatriz Lopez-Miranda ’99—Benjamin Wendel ’80—Berta Kvamme ’94—Bignesh Satpahty ’13—Bill Ketcham ’02—Bill
Prominski ’87—Bill Reid ’88—Bin S. Teh^ (NP) ’98—Blake Hocott (HMP) ’91—Bob Francis ’84—Bob Osborne ’85—Bob Stallworth
’92—Bradley J. Potsic (HMP) ’12—Brandon Langlinais ’08—Brandon Stroh ’05—Brian Carrier (HMP) ’11—Brian Ogden ’10—Brian
S. Haley ’14—Bruce Cheatham ’79—Bruce Morris ’97—Bryan L. VanZandt ’12—Bryan Prange ’09—Burton Spangler ’71—C.T. Morris,
Jr. ’80—Candy Roberts ’88—Carl E. Nuesch^ ’89—Carl Harrell ’81—Carlo A. Viamonte (NP) ’12—Carlos Farinas ’07—Carol Simmons
’68—Chad J. Goodman (NP) ’99—Chad M. Amosson^ ’01—Chad Mills ’02—Charles C. Trinh (HMP) ’99—Charles L. Moffet^ ’80—
Charlie Williams ’73—Charlotte Hayes ’93—Chitra Viswanathan (NP, HMP) ’05—Chris Canitz ’97—Chris Govea ’04—Christopher A.
Bedford ’17—Cindy Woo ’93—Colin Bray ’05—Colin Dodds (HMP) ’03—Corey Jensen (HMP) ’10—Craig Polson ’89—Craig Thiessen
’89—Daniel Bokhari ’17—Daniel Miller ’81— Daniel Reimer ’78—Daniel W. Fang (HMP)
’01—David A. Hughes• ’76—David Boals† ’69—David Diment ’89—David E. Goller ’98—
David Feldman ’93—David J. Sabbag ’16— David l. Irwin ’15—David Katz (NP) ’07—David
King ’87—David L. Janssen^ ’86—David S. Conrow (HMP) ’00—David Walker^ ’72—
David Wynne (HMP) ’11—Dean Chauvin (HMP) ’08—Derek Bergeson ’06—Don Evans ’07—
Donald Holmquest• ’74—Douglas Woo ’77—Doyle Simmons ’68—E. Gordon DePuey•
’77—Earljay O. Landrito ’17—Ebrahim S. Delpassand• ’90—Eduardo Martinez ’80—
Edward Callaway ’91—Edward Knudson ’74—Elan Omessi ’94—Elizabeth Koch ’85—
Elizabeth Rhea ’76—Elizabeth Weihe ’11—Ellis Eliza Deville^ ’77—Eric Trevino ’11—Ernesto
Blanco ’09—Errol Anderson ’85—Errol Candy ’89—Eugene Brown ’88—Ezequiel Silva ’01—
Farbod Nasseri (NP) ’13—Francisco Arraiza ’02—Frank Hadlock ’75—Frank Scalfano
’89—Fred Dean ’72—Fred Johnson ’77— Fred Quenzer ’72—G. Maria Rauch (HMP)
’08—Gardiner Bourque ’68—Garrett Anderson (HMP) ’02—Gary L. Horn ’15—Gary S. Likhari ’15—Gene Cunningham ’73—Gene
Smigocki ’87—George Boutros ’74—George Campbell† ’72—George Chacko• ’92—George Sofis ’05—George Soltes (HMP) ’94—
George Yama ’71—Gitesh Chheda ’09—Greg Boys (HMP) ’04—Gregory A. Patton^ ’81—Gregory H. Boering ’17—Guy Robitaille
’70—Hal Jayson ’83—Hana Khan ’11—Harry Butters ’78—Hollis Halford III ’81—Hsin H. Lu^ ’86—Hussain Thaver ’10—Iris W.
Gayed• ’97—Isabel Menendez ’85—J.P. Badami II ’83—J.P. Harris ’03—James Courtney (HMP) ’98—James D. Green ’70—James
E. Lefler ’98—James Fuchs ’80—James Lloyd ’83—James P. Caplan• ’81—James Tatum ’08—James P. Willis ’98—James Teague†
’78—James Van Dolah ’76—James W. Cole ’95—James Wolf ’91—Jamie Gregg ’09—Jamie Tsai (HMP) ’09—Jandra Kalus ’73—
Janine Mele ’03—Jason Salber (HMP) ’02—Javier Nazario ’09—Jeff Sheneman (HMP) ’06—Jeff Smith ’91—Jeffery Kass ’75—
Jennifer R. Cranny ’01—Jerry Polasek ’09—Jesus Castro-Sandoval^ ’83—Jett Brady ’97—Jimmy Hammond ’05—Joel Dunlap
’92—John Barkley ’04—John Baxt ’74—John Brooks ’69—John Clement (HMP) ’04—John Gundzik ’91—John H. Liem^ ’77—John
J. Nisbet (HMP) ’00—John K. Miller ’95—John L. Howard ’90—John Labis ’03—John Melvin ’83—John R. Bodenhamer ’96—John
Romero ’76—John S. Michels ’13—John Thomas ’93—John W. Wright ’90—John Wilbanks^ ’77—Jon Edwards ’78—Jorges
Gonzales ’74—Jose Guerra-Paz ’77—Jose L. Enriquez ’16—Jose V. Watson (HMP) ’01—Joseph Chan ’99—Joseph Schniederjan
’09—Joshua D. Kuban (HMP) ’14—Juan J. Ibarra ’10—Juliet Wendt• ’92—Kanchan Phalak (HMP) ’14—Karuna A Munjal ’14—Kapil
Shroff (HMP) ’11—Karen Simmons ’88—Karl Chiang ’91—Kathryn M. Lewis^ ’89—Keith Light• ’02—Kenny Q. Sam ’16—Keton Shah
’17—Kirsten Warhoe^ ’94—Kris Pun (HMP) ’08—Kunal Shah ’17—Kuri Farid ’78—Lakhwant Singh ’10—Larry Gaines^ ’72—Larry
S. Carpenter^ ’88—Larry Yeager ’79—Lawrence D. Hochman^ ’96—Lawrence J. Scharf^ ’88—Lee Shukl^ ’87—Leroy Halouska
’75—Lillian Orson ’89—Linda Ann Hayman ’76—Lorell Ruiz-Flores ’17—Lori Young ’70—M. Amir Ibrahim• ’97—Marc Siegel ’88—
Marc Willis ’07—Marcus Calderon• ’73—Maria Patino ’05—Marianne Greenbaum ’99—Mariano Nasser ’76—Marina Soosaipillai
’80—Mario Polo (HMP) ’09—Mark E. Augspurger^ ’00—Mark Matthews ’92—Mark Miller ’77—Mark Pfleger (HMP) ’91—Mark
Sokolay ’75—Mark Stallworth ’93—Martin Cain ’89—Mary Round ’92—Mathew J. Rose ’13—Mathew R. Galfione ’12—Matt Stair
’09—Matty C. Artunduaga ’15—Mehmet Gurgun (HMP) ’93—Mel Thompson ’96—Mian A. Ibrahim ’97—Michael A. Marks (HMP)
’90—Michael Cagan ’74—Michael D. Bastasch^ (NP) ’03—Michael Driver^ ’82—Michael G. Gunlock ’99—Michael H. Hayman^
’77—Michael Jaimes ’07—Michael Kerley^ ’92—Michael Maiers ’11—Michael Nguyen ’10—Michael T. Sinopoli^ ’03—Mike
Silberman ’91—Mike Sloan (MP) ’91—Mike Smith ’91—Milton Gray ’69—Modesto Sanchez-Torres (NP) ’97—Mohammad Khan
’13—Monica L. Huang ’99—Monte F. Zarlingo (HMP) ’00—Moran E. Telesmanich ’17—Munish Chawla ’97—Nan Garrett ’03—Nancy
Anderson ’84—Nathan S. Floyd^ ’02—Nathan W. Uy^ ’01—Naveen Bikkasani ’03—Neil Cooper ’86—Nicholas Kutka• ’71—Noah N.
Chason ’12—Nora A. Janjan^ ’84—Norman Harris ’74—Paul B. Horwitz (HMP) ’01—Paul Weatherall ’87—Pauline L. Bui ’14—Pedro
Diaz ’06—Pedro Gonzalez ’11—Pekka Ahoniemi ’73—Peter Feola ’95—Peter Kvamme• ’95—Peter N. Fata ’16—Phil Mihm (HMP)
’97—Philip Trover ’77—Philip Weaver ’74—Pieretta Ferro ’71—R. Cody Mayo (HMP) ’14—Rafael Aponte ’88—Rafael Vicens ’13—Raj
Cheruvu ’00—Raj R. Chinnappan ’13—Ralph Norton ’75—Ralph Sharman ’74—Ramesh
Dhekne• ’75—Ramin Naeini (NP, HMP) ’10— Raphiel Benjamin ’69—Raul Meoz^ ’76—Ravi
J. Amin (NP) ’00—Ray Somcio ’15—Ray Ziegler ’69—Reed A. Shankwiler ’90—Reese
James ’81—Richard A. McGahan^ ’90— Richard Fremaux ’81—Richard H. Oria ’90—
Richard Parvey ’80—Richard Wilson (HMP) ’94—Rishi Arya ’16—Robert A. Gilbert^ ’94—
Robert B. Poliner ’90—Robert Davidson• ’91—Robert Denman ’97—Robert Francis
’82—Robert J. Feiwell (HMP) ’95—Robert Malone ’88—Robert McLaurin, Jr.^ ’84—
Robert Sims ’85 Robert Sonnemaker• ’76— Robert Tien ’88—Roger Selouan (NP, HMP)
’11—Rogue Ferreyro ’78—Rohit Ramanathan ’16—Ron Brociner ’85—Ron Dillee ’73—
Ronald Eikenhorst ’79—Ross Zeanah ’85— Roy Longuet ’79—Rudy Garcia ’07—Russell D.
Putnam ’00—Russell W. Wright (HMP) ’12— Ryan Skinner ’10—Sachin Patel (HMP) ’13—
Sager Naik ’04—Salmann Ahmed ’06—Samer Harmoush ’16—Sami Juma ’84—Samuel
Serna ’07—Sanaz Javadi ’15—Sandra Hatch^ ’91—Sanjiv Gala ’06—Sara L. Moorhead ’17—Sarfarraz Sadruddin ’15—Sarina W.
Fordice ’98—Sasi Yallampall ’12—Sean Raj ’16—Sergy Lemeshko ’10—Scott Dorfman ’93—Scott Meshberger ’98—Scott S. Lenobel
’12—Scott Sandberg (HMP) ’92—Scott Y. Harada ’15—Sidney Roberts^ ’91—Sima C. Artinian ’98—Simon Trubek ’05—Sophia
Chatziioannou• ’96—Srini Malini ’77—Stanley D. Clarke^ ’91—Stanley Lim ’04—Stephen C. Greco^ ’97—Stephen C. Liaw ’16—
Stephen Long• ’81—Stephen Parven ’86—Stephen R. Lee ’14—Steve Hong (HMP) ’02—Steve Sax ’86—Steven J. DiLeo ’98—Steven
R. Rinehouse ’95—Steven S. Chua ’15—Steven Yevich ’13—Susan Fan ’89—Susan Pinero ’85—Susan Weathers ’81—Tamara Ortiz
’11—Tara A. Hagopian ’13—Tara Sagebiel ’06—Terry O’Connor ’96—Terry Reed ’72—Tessa Hudspeth ’17—Tex Clifford† ’68—Thanh
John Van^ ’02—Thomas Dumler ’79—Thomas Hedrick ’75—Thomas L. Atlas (HMP) ’96—Thomas Saadeh ’14—Tim Dziuk^ ’94—Tim
Seipel ’08—Tim Tsai (NP) ’96—Todd Lanier ’92—Todd Minken ’94—Todd Samuels ’79—Todd Tibbetts (NP, HMP) ’06—Tom Kaminski
’87—Tom Lepsch ’87—Tomas Jimenez ’08—Travis D. Lyons (HMP) ’13—Vasu Rao ’94—Vera Selinko ’85—Veral D. Amin ’17—
Veronica Lenge ’10—Vincenzo K. Wong ’15—Vivek Bansal (HMP) ’12—Vivek C. Yagnik (HMP) ’01—Vivek G. Sahani ’16—W. Sam
Dennis^ ’80—Warren H. Moore• ’82—Wassel H. Beal• ’74—Wendy S. Carpenter (HMP) ’99—William Cunningham ’83—William D.
Permenter, Jr.^ ’89—William Jordan ’70—William Kent ’80—William Rogers ’74—Wilmer Moran ’78—Xiao Shi ’17—Zachary Martin
’86—Zeke Silva (NP) ’01- Zeyad A. Metwalli (HMP) ’14
SCB

HMP, Honorary Medical Physicist; NP, Naresh Prasad Scholar; •, Nuclear Medicine;
^, Radiotherapy; †, R.I.P.

To my wife, Maria Thelma Salina-Clarke, who supported me through all the long nights of writing.
GDC


This edition of Magnetic Resonance Imaging is dedicated to my friends Janice and Robert McNair in
recognition of their incredible service to Houston and especially to Baylor College of Medicine. I have
been so fortunate to call them close friends these past 50 years and to witness their success and con-
tinuing philanthropy which amazes me and which includes substantial financial support for …

• AD Players • Houston Symphony

• Boy Scouts of the Sam Houston Council • Houston Zoo
• Columbia College • M.D. Anderson Cancer Center
• DePelchin Children’s Center • Menninger Clinic
• Discovery Green • Rice University
• Hermann Park Conservancy • San Jacinto Girl Scouts
• HISD Fine Arts Program • Texas Children’s Hospital
• Hope Lodge Houston • Texas Heart Institute
• Houston Baptist University • The Cotswold Project
• Houston Community College • University of South Carolina
• Houston Grand Opera • United Way Computer Training Center
• Houston Police Department

Their support for Baylor College of Medicine has been exceptional. The new hospital and clinic facili-
ties are located on the McNair Campus. Thank you my friends.

This page intentionally left blank
 PREFACE

When I began teaching/research at Baylor by an in-depth explanation of how MRI works

College of Medicine (BCM) in 1967 and a few (Parts III and IV). Parts V and VI of the text
years later when the Houston Community discuss the latest imaging methods. The final
College (HCC) was established, the subject section covers personnel and patient safety and
matter medical physics was a piece of cake. All I administration issues (Part VII).
had to teach were radiography and fluoroscopy. The American Board of Radiology (ABR) has
Today, all such students must know everything developed a certificate of Additional Qualifica-
known in 1967 plus all of the imaging modalities tions in MRI. The American Registry of Radio-
since, including MRI … in the same length of logic Technologists (ARRT) currently offers an
time! Now, the required fund of knowledge is advanced exam to qualify technologists in MRI.
ENORMOUS, and the demands on students are Both organizations will find that this text pre­
exceptional. sents the information needed for those advanced
In the early 1980s, magnetic resonance programs.
imaging (MRI) burst onto the scene as a diagnos- The American Board of Radiology also
tic imaging tool with even more intensity than includes substantial clinical questions involving
computed tomography (CT) had in the 1970s. Its MRI in its Core Examination and in the coming
similarities to CT are somewhat obvious, but the Certification Examination to be given first in
underlying physical principles are totally differ- 2018. Some fundamental knowledge of MRI
ent and challenging to imaging physicians and physics will be necessary for success by the can-
technologists alike. Whereas CT is an extension didate. The 40 RSNA/AAPM Web-based Instruc-
of x-ray imaging and the basic physics has been tionals currently include eight modules on MRI
well integrated into medical imaging training physics to be accessed by resident radiologists
programs, the physical basis for MRI has not and radiologic technologists alike.
been well integrated because it is fundamentally Kraig Emmert created the illustrations for
different. The technologist operating an MRI this edition. His talents and clever ideas added
system or the radiologist interpreting the MR humor and sense where concepts were sometimes
image must understand the basic principles of difficult to express.
this modality. Since the first edition of this book, published
This fourth edition presents the fundamentals in 1988, a wealth of advancements and changes
of conventional MRI and of the developing has occurred in this fast-moving field. This
fast imaging techniques of steady-state and accounts for the extensive revision throughout
transient-state gradient echo imaging. Concepts the book. I am especially indebted to Geoffrey
are presented in a way that is easily understood Clarke, PhD, University of Texas San Antonio,
by students, technologists, and physicians who for his contributions. As I am now late in
have little or no background in mathematics the fourth quarter of this game of life and trying
and physics. Interested readers will find a more to get to overtime, I am especially pleased
complex mathematical development in Appendix that Dr. Clarke is committed to continuing this
A: “The Bloch Equations.” project.
The text begins with an overview and an And I am pleased to repeat a statement of mine
introduction of the fundamentals of electricity first published decades ago in the third edition
and magnetism (Parts I and II). This is followed of “Radiologic Science for Technologists” because

ix
x PREFACE

it is more meaningful today for MRI. “No matter it to me, and you will be appropriately honored
what the task, people want to do it faster. Medical and memorialized.
Imaging is no exception.” This volume continues efforts to make medical
To test your skills, I have hidden an “error” imaging understandable and medical physics
in this fourth edition. Find this error and report fun.

Stewart Carlyle Bushong

 CONTENTS

PART I  Fundamentals, 1
1 An Overview of Magnetic Resonance Imaging, 2

2 Electricity and Magnetism, 17

3 Nuclear Magnetism, 39

4 Equilibrium–Saturation, 49

PART II  The Magnetic Resonance Image, 58

5 Radiofrequency Pulse Sequences, 59

6 Magnetic Resonance Imaging Tissue Parameters, 65

7 Manipulating Magnetic Resonance Image Contrast, 76

8 Fourier Transforms in Magnetic Resonance Imaging, 91

PART III  The Imaging System, 110

9 Magnetic Resonance Imaging Hardware, 111

10 Primary Magnetic Resonance Imaging Magnets, 127

11 Secondary Magnetic Resonance Imaging Magnets, 143

PART IV  Image Formation, 161

12 Digital Imaging, 162

13 A Walk Through the Spatial Frequency Domain, 180

14 The Musical Score, 193

15 Magnetic Resonance Images, 213

xi
xii CONTENTS

PART V  Pulse Sequences, 242

16 Spin Echo Imaging, 243

17 Chemical Shift and Magnetization Transfer, 263

18 Steady State Gradient Echo Imaging, 272

19 Hybrid Fast Imaging Techniques, 290

20 Echo-Planar Imaging, 298

PART VI  Applications, 304

21 Nuclear Magnetic Resonance Spectroscopy, 305

22 Partially Parallel Magnetic Resonance Imaging, 319

23 Magnetic Resonance Angiography, 331

24 Perfusion Imaging, 345

25 Diffusion Imaging, 354

26 Cardiac Magnetic Resonance Imaging, 360

PART VII  Safety, 374

27 Contrast Agents and Magnetic Resonance Imaging, 375

28 Magnetic Resonance Imaging Artifacts, 384

29 Biological Effects of Magnetic Resonance Imaging, 405

30 Managing a Magnetic Resonance Imaging System, 422

Appendix A The Bloch Equations, 442

Appendix B Additional Resources, 446
Practice Examinations, 449
Answers to Challenge Questions, 476
Answers to Practice Examinations, 489
Glossary of Magnetic Resonance Imaging Terms, 492
 PA RT
I
Fundamentals
 CHAPTER

1 
An Overview of Magnetic
Resonance Imaging

OBJECTIVES
At the completion of this chapter, the student • Distinguish between spatial resolution and
should be able to do the following: contrast resolution.
• Define nuclear magnetic resonance and • Discuss the symbols M, B0, Bx, FID, and T.
magnetic resonance imaging. • State the Larmor equation and discuss its
• Identify the three imaging windows of the significance.
electromagnetic spectrum.

OUTLINE
Historical Trail Multiplanar Imaging An Overview of Magnetic
Felix Bloch Magnetic Resonance Resonance Imaging
Damadian and Lauterbur Spectroscopy Net Magnetization
Why Magnetic Resonance No Ionizing Precession
Imaging? Radiation The Larmor Equation
Contrast Resolution Magnetic Resonance Imaging Free Induction Decay
Spatial Resolution Hardware Fourier Transformation

KEY TERMS
Contrast resolution Nuclear magnetism Spatial resolution
Electromagnetic radiation Precession Zeugmatography
Magnetic moment

HISTORICAL TRAIL
electromagnetic spectrum called the visible light
If someone wanted to make an image of a patient region (Figure 1-1).
150 years ago, what could have been done? Electromagnetic radiation can be character-
Actually, not much. At that time, only photogra- ized by any one of three parameters: energy,
phy or hand-drawn images were available. Both wavelength, and phase. The frequency is some-
types of such images use the narrow band of the times used to describe the wave character of

2
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 3

Energy Frequency Wavelength

(eV) (Hz) (m)
1010 1024 10−16
109 1023 10−15 Megavoltage therapy
108 1022 10−14 Gamma rays Supervoltage therapy
107 10−13 X- rays Diagnostic
1021 Contact therapy
1 MeV 106 1020 10−12
Grenz rays
X-ray 105 10−11
imaging 1019
104 1018 10−10
1 nm
1 keV 103 1017 10−9
102 1016 10−8 Ultraviolet (UV) Violet
101 10−7 Blue
Visual 1015 1 µm Visible light Green
imaging 1 eV 100 1014 10−6 Yellow
Infared (IR)
10−1 1013 10−5 Red
10−2 1012 10−4
10−3 1011 10−3
1 cm Microwaves
10−4 1010 10−2
10−5 109 10−1
1 GHz 1m
10−6 108 100
MR UHF
imaging 10−7 107 101
VHF
10−8 106 102 Shortwave
1 km Radiofrequency
10−9 1 MHz 105 103 (RF) Standard broadcast
10−10 104 Longwave
104
10−11 103 105
1 kHz
10−12 102 106

FIGURE 1-1  The electromagnetic spectrum showing values of energy, frequency, and wavelength for the imaging
windows of visible light, x-rays, and radio waves.

electromagnetic radiation but it is basically detected by a medium that is sensitive to that

equivalent to the wavelength since the wave- kind of radiation—a photographic emulsion or
length is just the speed divided by the frequency the retina. Therefore, a photograph is made with
and the speed of light is a constant, c (~3 × reflected electromagnetic radiation and a suitable
108 m/s). Although the only electromagnetic receptor.
radiation that we can sense directly is visible Nineteenth-century physicists studying visible
light, we know that the range of electromagnetic light detailed its wavelike properties according
radiation extends over many orders of magni- to how it interacted with matter (i.e., reflec-
tude and different types of radiation. tion, diffraction, and refraction). Consequently,
How does photography work? Visible light visible light has always been characterized by its
reflects off an object, and the reflected light is wavelength.
4 PART I  Fundamentals

Visible light extends from approximately 400 nm Commercial broadcasts such as AM radio,
(blue) to 700 nm (red). FM radio, and television (TV) are similarly iden-
tified. The AM RF band ranges from 540 to
1640 kHz, and the FM RF band ranges from 88
When Wilhelm Conrad Roentgen discovered to 108 MHz. TV broadcast ranges from 54 to
x-rays in 1895, there was suddenly another 806 MHz, which includes both VHF and UHF.
narrow region of the electromagnetic spectrum
from which medical images could be made. In Magnetic resonance images are made with RF in
1901 Roentgen received the first Nobel Prize in the range from approximately 10 to 300 MHz.
physics for his discovery. One reason Roentgen
received this award was that within 6 months, Use of the RF region of the electromagnetic
he had conducted a number of cleverly designed spectrum to produce an image is especially spec-
experiments and described x-rays much as they tacular. It is based on an analytical procedure
are known today. called nuclear magnetic resonance (NMR) and
Some of his experiments indicated that this was first called nuclear magnetic resonance
“x-light” interacted as a particle, not as a wave. imaging (NMRI). Some of the leaders in radiol-
As a result, x-ray emissions are identified accord- ogy were concerned about using the word nuclear
ing to their energy. Although we commonly refer around patients, since NMRI really didn’t involve
to kilovolt peak (kVp), it is more accurate to use any kind of ionizing radiation. As a result, that
kiloelectron volt (keV) to identify x-radiation. word was dropped early in the development of
this imaging process, and we are left with mag-
Diagnostic x-rays range from approximately netic resonance imaging (MRI).
20 keV to 150 keV. How is a magnetic resonance (MR) image
made? For a visible image, radiation is reflected
How is an x-ray image made? Electromag- from the body. For an x-ray image, radiation is
netic radiation (i.e., an x-ray beam) shines on transmitted through the body. For an MR image,
a patient. Some of the radiation is absorbed; the patient is stimulated so that electromagnetic
some of it is transmitted through the patient to radiation is emitted from the body. Through the
an image receptor. This results in a shadow gram- use of some clever methods, the emitted signal is
like image from the transmission of electromag- then detected, interpreted, and used to produce
netic radiation. an image (Figure 1-2).
During the latter part of the nineteenth century,
after Thomas Edison’s early work, engineers
Felix Bloch
and physicists worked to develop radio com­
munications. Electrons must oscillate in a con- Magnetic fields associated with atoms and nuclei
ductor to create a radio emission. This requires were first described in the 1930s. Otto Stern
the construction of an electronic circuit called and Isador Rabi each received a Nobel Prize in
an oscillator. The oscillator is the basis for physics for their work on atomic and nuclear
radioelectronics. magnetism. Rabi coined the term nuclear mag-
The electromagnetic radiation produced by netic resonance.
the oscillator is called a radiofrequency (RF) In 1946 Felix Bloch at Stanford and Edward
emission. Physicists identify this radiation accord- Purcell at Harvard independently described
ing to the frequency of oscillation. NMR in a solid. They shared the 1952 Nobel
Prize in physics for this work. Bloch continued
extensive studies with the NMR of water, thereby
RF radiation extends over a range from 3 kHz
laying the groundwork for later developments
to 3 GHz.
that led to MRI.
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 5

magnetic moment. Nucleons that have charge

(e.g., protons) and that spin have even stronger
magnetic fields.
Experimental verification for the Bloch equa-
tions did not come until the early 1950s. By 1960
several companies were producing analytical
instruments called NMR spectrometers. During
the 1960s and 1970s, NMR spectroscopy (see
Chapter 8) became widely used in academic and
industrial chemistry research. Such use of NMR
enabled investigators to determine the molecular
configuration of a material from the analysis of
its NMR spectrum.
A

Damadian and Lauterbur

In the late 1960s engineer-physician Raymond
Damadian, while working with NMR spec-
troscopy, showed that malignant tissue has a
different NMR spectrum from normal tissue.
Furthermore, he showed that the parameters
associated with NMR (i.e., proton density, spin-
lattice relaxation time, and spin-spin relaxation
B time) differ between normal and malignant
tissue. Damadian produced a crude NMR image
of a rat tumor in 1974 and the first body image
in 1976. That image took almost 4 hours to
produce.
At this same time Paul Lauterbur, an NMR
C chemist at State University of New York in
Stony Brook, developed the first imaging method
FIGURE 1-2  How images are made using the three regions
of the electromagnetic spectrum. A, Reflected visible light. using NMR that is similar to what is used today.
B, Transmitted x-rays. C, Emitted radiofrequency. He called this method zeugmatography, which
was sort of Greek for saying that this imaging
method requires a whole bunch of magnetic
fields whizzing and buzzing around. Meanwhile
Bloch is to MRI what Roentgen is to x-ray in Nottingham, England, Peter Mansfield, a
imaging, and Bloch is known as the father of solid-state physicist, was engaged in similar
MRI. As a theoretical physicist, Bloch proposed research and eventually developed the echo-
some novel properties for the atomic nucleus, planar MRI method that is used for functional
including that the nucleus behaves like a small MR neuroimaging today. In 2003 Lauterbur and
magnet. He described this nuclear magnetism by Mansfield shared the Nobel Prize in physiology
what are now called the Bloch equations (see and medicine “for their discoveries concerning
Appendix A). magnetic resonance imaging.”
Bloch’s equations explain that a nucleus, In fact, a large number of scientists made sig-
because it spins on an imaginary axis, has an nificant contributions to the early development
associated magnetic field. This field is called a of MRI, among them James Hutchinson and
6 PART I  Fundamentals

TA B L E 1 - 1 Approximate Spatial and Contrast Resolution Characteristics of Several

Medical Imaging Systems

Magnetic
Nuclear Computed Resonance
Medicine Ultrasound Radiography Tomography Imaging
Spatial resolution (mm) 5 2 0.05 0.25 0.25
Spatial resolution (lp/mm) 0.1 0.25 10 2 2
Contrast resolution (mm at 0.5% 20 10 10 4 1
difference)

William A. Edelstein at the University of Aber- Contrast resolution is the principal advantage
deen, David Hoult at Oxford University, Ian of MRI.
Young at EMI Laboratories, and Waldo Hinshaw
and E. Raymond Andrew, both at the University
Spatial Resolution
of Nottingham. These gentlemen have all received
numerous high scientific and engineering honors Spatial resolution refers to the ability to identify
for their contributions to the field. an object, usually a small, dense object like a
metal fragment or microcalcification, as separate
and distinct from another object. Table 1-1
WHY MAGNETIC shows representative values of spatial resolution
RESONANCE IMAGING? and contrast resolution for various medical
When a plain radiograph of the abdomen is imaging devices.
placed on a view box for interpretation, what In x-ray imaging, spatial resolution is princi-
can be seen? Not much. The image is gray and pally a function of the geometry of the system.
flat and shows little detail. A conventional tomo- Two important geometric considerations include
gram or an angiogram may be done to improve focal spot size and source-to-image receptor dis-
image contrast. tance (SID). In x-ray imaging, scatter radiation
limits the contrast resolution. X-ray beam colli-
mation and the use of radiographic grids reduce
Contrast Resolution scatter radiation and therefore improve contrast
If such an image is unsatisfactory, what else can resolution.
be done? A computed tomography (CT) image
can be requested. The principal advantage of CT CT has superior contrast resolution compared
imaging over radiographic imaging is superior to radiography because it uses a finely collimated
contrast resolution, the ability to image differ- x-ray beam, which results in reduced scatter
ences among low-contrast tissues. Contrast reso- radiation.
lution allows visualization of soft tissue with
similar characteristics, such as liver–spleen or In x-ray imaging, the x-ray attenuation
white matter–gray matter. coefficient (µ) determines the differential x-ray
The spatial resolution of a CT image is worse absorption in body tissues. In turn, the x-ray
than that of radiographic imaging because it is attenuation coefficient depends on the energy of
digital and limited by pixel size. Likewise, the the x-ray beam (E) and the atomic number (Z)
spatial resolution of MRI is worse than that of of the tissue being imaged.
radiography. However, the contrast resolution is The basis for the MR image is different.
even better with MRI than with CT. It is a function of several intrinsic NMR
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 7

characteristics of the tissue being imaged. The technique selection and positioning. Patient posi-
three most important tissue characteristics are tioning in radiography is important to ensure
proton density (PD), spin-lattice relaxation time that the structure being imaged is parallel and
(T1), and spin-spin relaxation time (T2). Second- close to the image receptor. MR images are
ary characteristics include flow, magnetic suscep- directly available as projections in any plane,
tibility, paramagnetism, and chemical shift. when the patient is properly positioned at the
There are two principal parameters to select magnet isocenter and with intended anatomy at
in the production of a radiographic image: kilo- the sensitive region of the RF coil.
volt peak (kVp) and milliampere-second (mAs).
By carefully selecting kVp and mAs, radiographer
Magnetic Resonance Spectroscopy
can optimize the contrast resolution of an image
without compromising the spatial resolution. Another advantage to MRI is the possibility
There are many parameters to select in the of doing in vivo magnetic resonance spectros-
production of an MR image. The time sequence copy (MRS). It is possible to make an MR
of energizing RF emissions (RF pulses) and gradi- image, see a suspicious lesion, put the cursor on
ent magnetic fields determines the contrast reso- that lesion, and encompass it within a region
lution. The principal pulse sequences are partial of interest (ROI). The radiologist then could
saturation, inversion recovery, spin echo, gradi- retrieve the NMR spectrum from that lesion for
ent echo, and echo planar. Each sequence has a analysis.
large selection of timing patterns for the RF Interpretation of the NMR spectrum could
pulses and gradient magnetic fields to optimize then tell whether the tissue is normal or abnor-
contrast resolution for visualization of various mal. If the tissue appears abnormal, the NMR
anatomical and disease states. spectrum could reveal the molecular nature of
the abnormality. Unfortunately, the chemicals
that we would like to study with MRS are at
Multiplanar Imaging concentrations thousands to tens of thousands
An additional advantage to MRI is the ability to of times less than the concentration of water
obtain direct transverse, sagittal, coronal, and in tissue. So MRS is performed only in very
oblique plane images. Conventional radiographs special cases where it is worth taking the extra
show superimposed anatomy regardless of the time.
plane of the image. In CT imaging, sagittal and
coronal images are reconstructed either from a Sensitivity describes how well an imaging system
set of contiguous images or directly from the can detect subtle differences in anatomy. Speci-
volumetric data of spiral CT. With MRI, a large ficity refers to the ability to precisely identify the
data set is acquired during a single imaging nature of such differences.
sequence from which any anatomical plane can
be reconstructed. MRI has excellent sensitivity. MR spectros-
Viewing images obtained from various ana- copy could provide increased specificity if there
tomical planes requires a different kind of knowl- were a way to get enough signal.
edge on the part of physicians and technologists.
Except for CT images, most x-ray images are
No Ionizing Radiation
parallel to the long axis of the body. The MRI
interpreter may view anatomical planes that have Another advantage of MRI over x-ray imaging is
not been imaged before. The required interpre- that MRI does not require ionizing radiation.
tive skills come with experience. This lack of ionizing radiation has been effec-
When students enroll in a radiologic tech­ tively used to promote the safety of MRI to the
nology program, the curriculum focuses on medical community and public.
8 PART I  Fundamentals

MRI does not require ionizing radiation. part to be imaged is at that position. An MRI
operating console has controls for selecting the
timing parameters, the field of view, and slice
thickness rather than kilovolt peak, milliampere,
MRI uses RF electromagnetic radiation and and exposure time.
magnetic fields, which do not cause ionization,
and therefore do not have the associated poten-
AN OVERVIEW OF MAGNETIC
tially harmful effects of ionizing radiation. Some
RESONANCE IMAGING
bioeffects of RF and magnetic fields are known
to exist, but the MRI systems are carefully The hydrogen nuclei in the patient, often just
designed to ensure that the levels reached are not referred to as protons, behave like tiny bar
high enough to cause harm and none of the bio- magnets. Hydrogen makes up 80% of all atoms
logical effects associated with MRI have been found in the human body, making hydrogen
linked to the induction of malignant disease. extremely useful for MRI. Because hydrogen is a
single-charged spinning nucleon, the hydrogen
nucleus exhibits magnetism due to its angular
MAGNETIC RESONANCE momentum and magnetic moment. Before the
IMAGING HARDWARE patient is put into the B0 magnetic field, the mag-
Just as a radiographic imaging system can be netic moments of the patient’s nuclei are ran-
identified by its main components (i.e., x-ray domly oriented (Figure 1-3).
tube, high-voltage generator, and operating The small arrows in Figure 1-3 represent these
console) so can an MRI system be identified by individual proton magnetic moments, which are
its main components (i.e., magnet, computer, and also referred to as magnetic dipoles. Under
operating console). normal circumstances, these magnetic dipoles
The magnet is typically a large cylindrical (each has a north and south magnetic pole) are
device that accommodates the patient during randomly distributed in space. Consequently, if
imaging. Unlike a CT gantry, the MRI magnet the net magnetic field of a patient were measured,
does not have moving parts. The only things it would be zero because all of the individual
that move are electrons in a conductor. The magnetic dipole moments cancel.
patient aperture is usually approximately 60 cm
in diameter. RF coils surround the patient in this
Net Magnetization
aperture. Gradient coils, shim coils, and, in the
case of an electromagnet, primary coils all sur- When the patient is placed in the presence of a
round the RF coils to produce the required mag- strong external magnetic field, some of the indi-
netic fields. vidual nuclear magnetic moments align with the
The computer required for MRI is similar to external magnetic field (Figure 1-4). The Carte-
that used for CT: very fast and with high capac- sian coordinate axis, X, Y, and Z, is always ren-
ity. During an MRI examination, more data are dered with the Z-axis as the vertical axis as
collected, and the computations required are
longer and more difficult than those for CT.
The MRI operating console is also similar to
that used for CT in that it has the same controls
for postprocessing and annotation of an image.
A CT system uses mechanical incrementation for
patient localization and so does an MRI system. FIGURE 1-3  Under normal conditions, nuclear magnetic
The patient undergoing MRI is moved to the dipoles in the body are randomly distributed, which results
isocenter of the magnet to ensure that the body in zero net magnetization.
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 9

Y
X B0
Z M

Z M

FIGURE 1-4  When a strong external magnetic field (B0) is applied, the patient becomes polarized and has net magne-
tization (M).

shown. Vector diagrams that show this coordi-

nate system will be used to develop the physics
of MRI.
Precession
A vector is a quantity that has magnitude and
direction.
Gravity
The Z-axis in Figure 1-4 is drawn along
the long axis of the patient. By convention in
MRI, the Z-axis coincides with the axis of the
static magnetic field (B0). In superconducting MR
imaging systems, the static magnetic field (B0) is
usually horizontal, making the Z-axis horizontal,
too. In a permanent magnet imaging system, the
Z-axis is usually vertical, as in the vector dia- FIGURE 1-5  When spinning in outer space, a gyroscope
grams that follow. just spins. On the Earth, however, it precesses as it spins.
The strength of the B0 magnetic field is
expressed in units called tesla (T). A frequent
MRI question is, “How many gray equal 1
Precession
tesla?” The answer is none. A relationship does
not exist between ionizing radiation and mag- In addition to polarization, another phenomenon
netic field strength. Nothing about MRI can be occurs when a patient is placed in a static mag-
measured in gray because MRI does not use ion- netic field. This phenomenon can be understood
izing radiation. by considering the gyroscope. A gyroscope has
With a patient positioned in a static magnetic an annulus of heavy metal attached by spokes to
field, proton dipoles align with the (B0) magnetic an axis (Figure 1-5). If the gyroscope is taken
field (see Figure 1-4). This statement is an over- into space and spins, it only spins. However, if
simplification and is not entirely true. Only the gyroscope spins on Earth in the presence of
one of approximately every million dipoles a gravitational field, not only will the gyroscope
becomes so aligned. However, once aligned, the spin, but it will also wobble. Physicists call this
patient is polarized and has a net magnetization. wobbling motion precession.
The patient now has a north and a south mag- Precession is the interaction between the spin-
netic pole. ning mass of the gyroscope and the mass of the
10 PART I  Fundamentals

FIGURE 1-6  The International Space Station. (Courtesy National Aeronautics and Space Administration.)

Earth that is manifest through the gravitational B0

field. By spinning, the gyroscope creates angular
momentum, which interacts with the angular
momentum of the spinning Earth and causes the
precessional motion.
Early space station designs were gyroscope-
like saucers, spinning to produce an artificial
gravity. Currently, hand and foot clips provide
moorings for inhabitants of the International
Space Station (Figure 1-6).
FIGURE 1-7  In the presence of an external magnetic field
(B0), a spinning proton precesses.
The gyroscope precesses in the presence of
gravity; the proton precesses in the presence
of B0.
The Larmor Equation
Similarly, if a spinning magnetic field, such as
the magnetic moment of the proton (Figure 1-7), The following is the fundamental equation for
is in the presence of a static magnetic field, it will MRI, the Larmor equation. This equation identi-
not only spin but will also precess. fies the frequency of precession.
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 11

Larmor Frequency TABLE 1-2 Nuclei of Medical Interest

fo = g Bo and Their Gyromagnetic
where fo is the frequency of precession (or reso- Ratios
nant freqeuncy) and γ is the gyromagnetic ratio.
Nucleus Gyromagnetic Ratio (MHz/T)
1
H 42.6
The Larmor equation relates B0, the strength 19
F 40.1
of the static magnetic field, to the precessional 31
P 17.2
frequency (f) through the gyromagnetic ratio (γ), 23
Na 11.3
which has a precise value characteristic of each 13
C 10.7
nuclear species. Sometimes you will see the gyro- 2
H 6.5
magnetic ratio expressed as γ, but this version is
17
O 5.8
39
to be used only with angular frequencies scaled K 2.0
as radians/seconds.

g =g
2p
The symbol MZ represents net magnetization
is the correct version of the gyromagnetic ratio that lies along the Z-axis.
to use with frequencies scaled in Hz. The MRI experiment begins with the emission
Gyromagnetic ratio is to MRI what the disin- of a pulse of RF energy at the Larmor frequency
tegration constant is to radioactive decay. Each from an inductor, called an RF coil, into the
radionuclide has its own characteristic disinte- patient (see Figure 1-8). For hydrogen imaging
gration constant; each nuclear species has its with a magnetic field of 1 T, the RF coil is tuned
own characteristic gyromagnetic ratio. to 42 MHz.
The units of the gyromagnetic ratio, scaled as If one plucks a string of a guitar and a harp
described above, are megahertz per tesla. For is standing nearby, one of the strings on the harp
example, hydrogen has a gyromagnetic ratio of will begin to vibrate (Figure 1-9); the other
approximately 42 MHz/T. If B0 is 1 T, then the strings will remain still. The harp string vibrates
precessional frequency is 42 MHz. Likewise, at because that string has the same fundamental
1.5 T the precessional frequency is 63 MHz. The resonance as the plucked guitar string. The “R”
precessional frequency is also called the Larmor in MRI stands for resonance. The RF pulse trans-
frequency. mitted into the body must be at the resonant
Table 1-2 shows the principal nuclei of bio- frequency of the precessing hydrogen nuclei for
logical interest in MRI. Medical applications of energy to be transferred and imaging to occur.
MRI concentrate on hydrogen because of its rela- Most objects in nature have a fundamental
tive abundance and high gyromagnetic ratio. resonance. Energy transfer is always most effi-
Compared with other nuclei in the body, hydro- cient at resonance. For example, at a large hotel
gen is the best for producing an MR signal. in Kansas City several years ago, people were
dancing on a suspended bridgelike walkway.
They hit a resonance that was fundamental to
Free Induction Decay the walkway. The walkway collapsed, killing
When a patient is placed in the B0 magnetic field, several people. For this reason, marching mili-
the patient becomes polarized (see Figure 1-4). tary personnel are instructed to break cadence
The proton magnetic dipoles have aligned with when crossing a bridge. A third example is the
B0, and the alignment is symbolized with one collapse of the Tacoma Narrows suspension
large arrow, MZ (Figure 1-8). This arrow repre- bridge when it was subjected to harmonic buf-
sents a vector quantity called net magnetization. feting winds (Figure 1-10).
12 PART I  Fundamentals

Y X
B0

MZ

FIGURE 1-8  Net magnetization along the Z-axis RF coil

is represented by MZ and the large arrow.

When RF is pulsed into the patient, the protons

individually flip and give up their energy to the
patient while continuing to precess. Then, as a
group, the net magnetization grows to its normal
state in the positive Z direction. The normal
state is called the equilibrium magnetization state
because the protons are at equilibrium in the B0
magnetic field. As the individual protons return
to equilibrium, the net magnetization precesses
around the Z-axis and slowly returns (relaxes)
back toward equilibrium (Figure 1-13).
To a disinterested observer, such as the RF
FIGURE 1-9  Plucking one guitar string causes only one receiving coil shown in Figure 1-13, such preces-
string of a nearby harp, which has the same fundamental sion is not obvious. Only a magnetic field that
resonance, to vibrate. first approaches and then recedes harmonically
is observed.
With any moving magnetic field, an electric
current can be induced in a properly designed
With net magnetization in the Z direction, not coil. The induced current represents a radio
only are the proton magnetic dipoles aligned, but signal emitted by the net magnetization created
each individual proton is precessing at the Larmor by the nuclei in the patient. This signal is called
frequency (Figure 1-11). When the RF signal is a free induction decay (FID). The RF coil sur-
pulsed at resonance into the patient, the energy rounding the patient receives an oscillating signal
state of many protons is changed. The net mag- that decreases with time (Figure 1-14). The signal
netization, due to all of the protons, is said to decreases with time as the proton spins begin to
“flip” toward the negative Z direction, while lose phase coherence or dephase.
still precessing about the Z-axis (Figure 1-12). The time constant that describes this process
This precession is always perpendicular to Z, is known as a relaxation time, specifically T2,
in the XY plane, and if initially all of the spins which is also called the transverse relaxation
are aligned along the same direction in the XY time. It is similar to the decay constant that
plane, we have created a condition called phase describes radioactive decay. Two such relaxation
coherence of the spins. This is the condition in times exist in MRI. The other is the T1 relaxation
which the most MR signal can be generated and time that describes the rate of the magnetization
received. increasing back to equilibrium. T1 and T2 can
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 13

FIGURE 1-10  The Tacoma Narrows suspension bridge collapsed in buffeting gale force winds that set up a resonant
oscillation. (Courtesy Civil Engineering Department, Rice University, Houston.)

Y
X
B0

FIGURE 1-11  Placing a patient in a magnetic

field (B0) polarizes the patient and causes each
proton dipole to precess randomly.
14 PART I  Fundamentals

Relaxation time
Y
X RF
signal
intensity
Time
Z

RF transmit
FIGURE 1-14  The free induction decay is a decreasing
harmonic oscillation of the Larmor frequency.
FIGURE 1-12  Net magnetization changes along the Z
direction and the protons precess in phase when a proper
radiofrequency (RF) pulse is transmitted into the patient.
intensity versus frequency. Each of the peaks in
the NMR spectrum represents one characteristic
of the tissue under investigation.
RF receive coil How is an image obtained from an NMR
spectrum? The following is a simplistic explana-
Y RF signal
X tion. Figure 1-16 presents a transverse cross-
section through the trunk of the body. The
patient lies in a uniform B0, and two pixels are
highlighted. If both pixels contain the same
Z
tissue, the peak in the NMR spectrum represents
both pixels. One can tell by looking at the
spectrum what is in both pixels but cannot deter-
RF transmit coil mine how much of the signal comes from each
location.
FIGURE 1-13  Precessing net magnetization induces a If in addition to the uniform (B0) magnetic
radiofrequency (RF) signal in a receiving coil. That RF signal field, a gradient magnetic field (Bx) is superim-
is called a free induction decay.
posed across the patient that varies in field
strength, spatial localization is possible (Figure
1-17). The magnetic field will then change with
generally be considered to be independent of one the x-position, Btotal = B0 + BX ⋅ x.
another and represent two different processes The position where x = 0 is called the isocenter
occurring at the same time but often at different of the magnet, and here the gradient will be zero.
rates. However, as the position is moved away from
isocenter, the magnetic field will be increased or
diminished by an amount equal to +BX ⋅ x. As a
Fourier Transformation result, even though they represent the same
The FID is a plot of MR signal intensity as a tissue, the tissue in the pixel at the lower mag-
function of time (see Figure 1-14). If a mathemat- netic field strength has a lower Larmor frequency
ical operation called a Fourier transformation than the one located at the higher magnetic field.
(FT) is performed on the FID, the result appears The FID in this situation is considerably more
as an NMR spectrum (Figure 1-15). complex. After FT, this spectrum has two peaks
Whereas the FID is a graph of signal intensity instead of one. These two peaks carry spatial
versus time, the NMR spectrum is a graph of information. One represents the pixel at the
signal intensity versus inverse time (s−1), or hertz lower magnetic field; the other represents the
(Hz). Therefore the NMR spectrum is signal pixel at the higher magnetic field.
 CHAPTER 1  An Overview of Magnetic Resonance Imaging 15

Signal
RF intensity
signal
intensity
FT

Time(s) Frequency (s−1)

FIGURE 1-15  When a Fourier transformation (FT) is performed on the free induction decay, a nuclear magnetic reso-
nance spectrum results.

FIGURE 1-16  If the same tissue were in the two highlighted pixels, both pixels would be represented by the same peak
in the nuclear magnetic resonance spectrum.

FIGURE 1-17  In the presence of a gradient magnetic field, BX, the nuclear magnetic resonance spectrum provides
information on pixel location.
16 PART I  Fundamentals

application of gradient magnetic fields superim-

posed on the B0; however, the reconstruction
of an image occurs through a process called
two-dimensional Fourier transformation (2DFT)
or three-dimensional Fourier transformation
(3DFT). This is a special application of higher
mathematics that will be developed conceptually
later.

CHALLENGE QUESTIONS
1. What three windows in the electromagnetic
spectrum are available for the production of
medical images?
FIGURE 1-18  Projections can be obtained by rotating 2. What type of energy is involved in making
the gradient magnetic field around a patient. An  MR images, and how would you describe
image can be reconstructed from these projections by
that energy?
backprojection.
3. What are the three fundamental properties
of an electromagnetic wave?
4. What is contrast resolution?
A uniform magnetic field is required for 5. What is spatial resolution, and why is it
NMR spectroscopy; gradient magnetic fields are important in medical imaging?
required for MRI. 6. Why does MRI exhibit superior contrast
resolution?
Multiple projections can be obtained in MRI 7. The terms sensitivity and specificity are fre-
by electronically rotating the gradient magnetic quently used to describe imaging modalities.
fields around the patient to produce a set of How do they differ?
projections (Figure 1-18). The projections are 8. State the Larmor equation and identify each
Fourier transformed and then back projection parameter.
reconstruction can be used to produce an image 9. The description of the physical basis for MR
as in CT. Paul Lauterbur’s early MR images were imaging relies heavily on vector diagrams.
created in this manner. What is a vector diagram?
MR images are reconstructed differently now. 10. Why does a toy top wobble when spun, and
The spatial information still comes from the what is that wobble called?
 CHAPTER

2 
Electricity and Magnetism

OBJECTIVES
At the completion of this chapter, the student • Describe the nature of magnetism.
should be able to do the following: • Distinguish between magnetic susceptibility
• Discuss the roles of the following: Franklin, and magnetic permeability.
Coulomb, Tesla, Faraday, and Oersted. • Examine the role of electromagnetism in
• Describe the nature of the electric field. magnetic resonance imaging (MRI).
• List the four laws of electrostatics. • Define imaging window.

OUTLINE
Electrostatics Magnetism Electromagnetic Radiation
The Coulomb Magnetic Domains Maxwell’s Wave Equations
Electrification The Laws of Magnetism The Electromagnetic
The Electric Field The Magnetic Field Spectrum
Electrostatic Laws Electromagnetism
Electric Potential Energy Oersted’s Experiment
Electrodynamics Faraday’s Law
Electric Current
Ohm’s Law
Electric Power

KEY TERMS
Cryogenic Energy source Semiconductors
Energy sink Insulators Superconductor

Both electric and magnetic fields and electromag- fields interact with the intrinsic nuclear magne-
netic radiation are used in magnetic resonance tism of tissue, the proton dipoles. The tissue is
imaging (MRI). These are not unlike the physical excited with a radiofrequency (RF) pulse pro-
agents used in x-ray imaging, although their duced by an electrically stimulated coil, which
method of use is vastly different. usually also receives the magnetic resonance
Electricity is used to produce the primary (MR) signal emitted from the body.
static magnetic field (B0) and the secondary gra- A modest understanding of the fundamental
dient magnetic fields (BX, BY, BZ). These magnetic physical concepts of electricity and magnetism is

17
18 PART I  Fundamentals

necessary for an understanding of MRI. A logical

sequence for dealing with such subjects is to
discuss electrostatics and electrodynamics to
develop an understanding of electricity. This is
followed by a discussion of the phenomenon of
magnetism, which leads into electromagnetism
and electromagnetic radiation.

ELECTROSTATICS
The smallest unit of electric charge is contained
in the electron, and it is negative. The proton
likewise contains one unit of positive electric
charge. Electric charge, unlike other fundamen-
tal properties of matter such as mass, cannot
be subdivided. Furthermore, larger quantities
of charge can only be multiples of the unit FIGURE 2-1  Benjamin Franklin described electricity as the
charge. flow of positive electrification rather than electrons.
Although the magnitude of the electric charge
of an electron and proton is the same, the mass
of the proton is approximately 1840 times the
mass of the electron. Protons are relatively fixed
by virtue of their position in the nucleus of an identified as the fundamental charged particle
atom, whereas electrons are free to migrate from responsible for electricity.
atom to atom under some circumstances.
The Coulomb
Electrostatics deals with stationary electric
charges. The unit of electric charge is the coulomb (C),
with 1 C consisting of 6.24 × 1018 electronic
It was not until the 1750s that Benjamin charges, a sizable number of electrons. This defi-
Franklin first described the nature of electric nition, adopted in 1910, is such a strange number
charge. Franklin’s experiments have been popu- because the system for electrical measurement
larly associated with flying kites, which is indeed had already been established before the discovery
true (Figure 2-1). However, he was and is also of the electron. Ideally, the electron should be the
credited with being a laboratory scientist. Nev- smallest unit of electric charge. Instead, the
ertheless, Franklin erroneously assumed that charge on an electron is 1.6 × 10−19 C.
positive charges were migrating down his kite
string. 1 C = 6.24 × 1018 electrons; 1 electron = 1.6 ×
Franklin called this migration of charge elec- 10−19 C.
tricity and therein lies the origin of a confusing
convention: that electric current (I) in a conduc- The coulomb is an unfamiliar quantity to
tor flows opposite to electron movement. most people. The lightning associated with the
Franklin is also credited with much of the current thunderstorm shown in Figure 2-2 ranges from
terminology dealing with electricity: charge, dis- perhaps 10 to 50 C. The shock experienced when
charge, battery, shock, positive, negative, plus, grasping a doorknob in the dry air of winter is
and minus among others. Not until the work of measured in microcoulombs (µC), a mere 1012
J.J. Thompson in the 1890s was the electron electrons.
 CHAPTER 2  Electricity and Magnetism 19

FIGURE 2-2  Lightning is the migration of electrostatic charge, usually from cloud to cloud but also from cloud to ground.

Electrification
object so that the electrons are transferred by
Whenever electrons are added to or removed spark. Lightning bolts jump from cloud to cloud
from material, the material is said to be electri- or cloud to earth to shed themselves of excess
fied. Electrification can occur by contact, fric- electrons. The earth is the ultimate sink for excess
tion, or induction. electrons and is called an earth ground or just
Electrification by contact occurs when an ground in engineering terms.
object having an excess number of electrons
contacts a neutrally charged object or an object
The Electric Field
with a deficiency of electrons. Rubbing a balloon
over your hair to make it stick to the wall is an Charge cannot be destroyed by conversion to
example of electrification by friction. The loosely another form. In the universe, the total number
bound electrons of hair are mechanically trans- of negative charges equals the total number of
ferred so that the balloon becomes electrified. positive charges, and the net charge is always the
same—zero. Furthermore, charge is quantized; it
Induction refers to the transfer of mass or energy comes in discrete bundles rather than in a con-
between objects without actual contact between tinuum of values.
the objects. A force field called the electric field (E) is
associated with each charge. The proof of the
Electrification by induction occurs when a electric field is much the same as that of a gravi-
highly electrified object comes close to a neutral tational field—the exertion of a force.
20 PART I  Fundamentals

In an electric field the lines of force begin on

positive charges and end on negative charges.
+ –

Alternatively, if the positive charge is in the

A B field of a proton, the positive charge will be
repelled, and the imaginary lines will radiate
from the proton (see Figure 2-3). Neutral objects,
such as a neutron, do not have an electric field.
+ + – – The magnitude of the electric field is defined
as the force on a unit charge in the field as
follows:
C D
Electric Field
E = F/Q
+ – N N where E is the electric field intensity (newton/
coulomb), F is the force on the charge (newton),
and Q is the electric charge (coulomb).
E F
FIGURE 2-3  Electric fields radiate out from a positive Although electric charge is discrete, its associ-
charge (A) toward a negative charge (B). Like charges ated electric field is continuous, and this is a
repel one another (C and D). Unlike charges attract one fundamental characteristic of field theory. Parti-
another (E). Uncharged particles do not have an electric
field (F).
cles add in a quantized fashion; fields add in a
continuum by superposition. For example, the
force on an electric charge is determined by
The gravitational field produces a force that not only the movement of that charge but
causes one mass to be attracted to another. Simi- also the electric fields produced by all other
larly, the electric field creates a force between one electric charges near and far. That is field
charge and another. Although the gravitational superposition.
force is always attractive, the force of the electric
field can be attractive or repulsive, depending on
Electrostatic Laws
the nature of the charges involved.
The electric field is most easily visualized as The nature and intensity of the electric field
imaginary lines radiating from an electric charge form the basis for the four principal laws of
(Figure 2-3). The intensity of the electric field is electrostatics.
proportional to the concentration of lines, and Unlike versus Like Charges.  Because of the
therefore the electric field decreases as the square vector nature of the electric field, like electric
of the distance from the charge. charges repel, and unlike electric charges attract.
The electric field is a vector quantity; that is, Particles having no electric charge, such as neu-
it has not only magnitude but also direction. The trons, are not influenced by an electric field.
direction of the electric field is determined by the Coulomb’s Law.  The force of attraction or
movement of a positive charge in the electric repulsion between electrostatic charges was first
field. In the presence of an electron, the positive described by Charles Coulomb in the 1780s.
charge is attracted, and therefore the imaginary Coulomb noted that the force was proportional
lines of the electric field are directed toward the to the product of the two charges and inversely
electron. proportional to the square of the distance
 CHAPTER 2  Electricity and Magnetism 21

separating them. Coulomb’s law can be stated as Free electrons are distributed on the surface of
follows: the object, not inside it.

Coulomb’s Law
Charge Concentration.  If the electrified object
E = F/q = k Q/d 2 is regularly shaped, such as a sphere or wire, the
or distribution of electrons on the surface will be
F = k q Q/d 2 uniform. On the other hand, if the surface is
where F = the force (newton) irregularly shaped (for example, has a point as
q and Q = the charges (coulomb) in an electrified cattle prod or a microfocus field-
  d = the distance between them (meter) emission x-ray tube) electrons will be concen-
  k = a constant trated at the sharpest region of curvature.

The electrostatic force is one of the five fun-

Electric Potential Energy
damental forces in nature. It is 1037 times stron-
ger than the gravitational force, 1011 times as When two like charges are pushed together or
strong as the weak interaction, about equal to two unlike charges are pulled apart, work is
the magnetic force, and about 1137 of the strong required. The resulting system has the ability to
nuclear force. do work and therefore has electric potential
An electrified balloon will attract paper or energy.
repel a small stream of water. The electrostatic The work done to create electric potential
force of attraction holds electrons in orbit in an energy comes from an energy source, and the
atom. The electrostatic force of repulsion among work obtained from the potential energy is
protons is countered by the attraction between deposited in an energy sink. A hydroelectric
neutrons and protons by the strong nuclear force generator and a steam generator are common
but ultimately limits the size of the nucleus. energy sources that convert mechanical energy
Charge Distribution.  Because protons are fixed, into electric potential energy. Energy sinks, such
whereas electrons are free to move, the remaining as motors, lamps, and heaters, abound.
discussion of electrostatic phenomenon concerns The electric potential energy of a charge, q, is
electrons. Because the electric field associated converted to kinetic energy when the charge is in
with each electron radiates uniformly, electrons motion and work is done. The electric potential
on any electrified object tend to be separated energy of a charge influenced by the electric field
uniformly and to the maximum dimensions pos- of other charges is given by the following:
sible (Figure 2-4).
Electric Potential Energy
E = QV
or
W = QV
or
V = W/Q
where E is the potential energy (joule), Q is
the electric charge (coulomb), V is the electric
potential (volt), and W is work (joule).

FIGURE 2-4  Electrons are distributed on the surface of Electric potential has units of joule per
electrified objects. coulomb, and it is not a force at all but rather a
22 PART I  Fundamentals

FIGURE 2-5  This Houston freeway at rush hour is analogous to an electric current, with the cars serving as electrons.

force used to do work. Therefore electric poten- there are fewer cars in each route, but the total
tial is given a special name, volt, and is com- flow will remain constant.
monly called voltage.
Electric Current
ELECTRODYNAMICS The electrons flowing in a conductor behave like
Electrodynamics involves what is commonly the automobiles on the freeway. For an electric
called an electric current or electricity. Electricity current to exist, a closed circuit is necessary. Each
deals with the flow of electrons in a conductor. electron must have a place to go. If there is a
barricade in a conductor, such as an open switch,
The science of electric charge in motion is electron flow ceases.
electrodynamics. The reason for investigating electricity is that
work can be extracted from the kinetic energy of
A Houston freeway at rush hour is analogous electrons moving in a conductor. The number of
to an electric current (Figure 2-5). Normally, electrons involved is given a special name, the
about 10,000 cars (electrons) each hour will pass ampere (A), which is equal to the flow of 1 C
any given point on a six-lane freeway. If there is each second (1 A = 1 C/s).
a wreck or construction, the speed of each car is An ampere is a rather large electron flow,
reduced (resistance) at that point and the flow of 6.24 × 1018 electrons each second. Electrons
traffic restricted. At an interchange, some cars cannot be counted that fast, so electric current is
may exit onto alternative routes. This allows measured by the associated magnetic field.
those cars remaining on the main freeway and Electric currents range from thousands of
those that exited to travel faster (parallel circuit). amperes in lightning bolts to picoamperes in elec-
The traffic flow in each branch is reduced because tronic equipment. Household current can be up
 CHAPTER 2  Electricity and Magnetism 23

Battery

Open

Close

Current (A)

Switch closed
Starting motor
0 Time
Switch
open
FIGURE 2-6  The electric circuit in an automobile is an example of direct current.
Current (A)

1/120 s
0 Time

Switch Switch
open closed

FIGURE 2-7  Normal household current is provided as an alternating current.

to approximately 30 A on any circuit. A current the switch is open, no current flows. When the
of only 100 milliamperes (mA) at 110 volts is switch is closed, electrons first flow in one direc-
almost always fatal, which is the reason grounded tion and then reverse direction and flow in the
circuits and ground fault circuit interrupters are opposite direction.
necessary elements in home wiring. When electrons begin flowing in the positive
Direct Current.  If the energy source propels direction, they begin slowly and speed up to a
electrons in only one direction, as with an auto- maximum, represented by the first peak of the
mobile battery, the form of electricity is direct current waveform (see Figure 2-7). Then they
current (DC) (Figure 2-6). At time zero, when the begin to slow down, still traveling in the same
switch is open, no current flows. The instant the direction, until they momentarily come to rest.
switch is closed, electrons flow—but only in one This moment of rest occurs 120 times each
direction. second and is represented by the zero crossing of
Alternating Current.  On the other hand, if the the waveform. Then the electrons reverse direc-
energy source is of an alternating form (Figure tion, first speeding up to a maximum, then
2-7), alternating current (AC) is produced. When slowing down to zero again.
24 PART I  Fundamentals

Such heating may be a problem with a resis-

Single tive electromagnet-type MRI system. The heating
phase can require a closed cooling system incorporat-
ing a high-efficiency heat exchanger. Supercon-
ducting MRI magnets do not have this difficulty
because the resistance-to-electron flow is zero.
The electrons are basically walking along empty
Three
phase
sidewalks.
The resistance of a conductor or circuit
element is usually a fixed quantity and follows
FIGURE 2-8  Electric power is generated and transmitted a simple relationship first described by George
in three-phase form.
Ohm in the 1840s.

The number of electrons in the circuit remains Ohm’s Law

constant, and their net movement is zero. Both
AC and DC are used to great advantage in MRI. R = V/I
Phase.  The current waveform shown in Figure where R is the resistance in ohm (Ω), V is the
2-7 illustrates that at any instant, all electrons electric potential in volt (V), and I is the electric
are moving in the same direction with the same current in ampere (A).
velocity. They are said to be in phase, and
the electricity is commonly called single-phase Although Ohm’s law is fundamental to elec-
current. tronics, many electric devices are used because
Actually, commercial electric power is gener- they do not obey Ohm’s law. The integrated cir-
ated and transmitted as three-phase current. A cuits of computer chips are prime examples.
three-phase waveform is shown in Figure 2-8. At Materials used in electric circuits are some-
any instant, not all electrons are moving in the times classified by their resistance. Those with
same direction, and those that are have different low resistance, such as copper, aluminum, and
velocities. They are out of phase with one another. seawater, are called conductors. Those with high
The concept of phase is important to the under- resistance, such as quartz, rubber, and glass, are
standing of the way an MR image is produced. called insulators.
Some materials that lie between conductors
and insulators are called semiconductors. Silicon,
Ohm’s Law selenium, and germanium are semiconductors
Electrons do not flow unimpeded in a circuit. used extensively in fabricating diodes, radiation
They behave much like an individual walking detectors, and all types of computer chips.
along a crowded sidewalk, bumping into people. If the electric resistance of a material is zero,
Electrons bump into other electrons of the con- that material is a superconductor. However, to
ductor. This property of any electric device is behave as a superconductor, as niobium and tita-
called impedance, and it is a function of the size, nium do, a material must be in an extremely cold
shape, and composition of the conductor or or cryogenic environment. This electric classifica-
circuit element. tion of material is summarized in Table 2-1.
There are three types of electric impedances:
capacitive, inductive, and resistive. If the work
Electric Power
done on the device changes the electric energy
into heat, the impedance is resistive, and it is When an electric current flows, it will do so
equal to the electric potential divided by the because of an electric potential (volt). Because of
current. the impedance of circuit elements to electron
 CHAPTER 2  Electricity and Magnetism 25

TA B L E 2 - 1 The Four Electrical States

of Matter

Class Property Material

Insulator Resists the flow of Rubber, glass,
electrons plastic A B
Semiconductor Can behave as an Silicon, FIGURE 2-9  A, In most matter, magnetic domains are
insulator or a germanium randomly oriented. B, Magnets exist when magnetic
conductor domains are aligned.
Conductor Allows the flow of Copper,
electrons with aluminum
difficulty
Superconductor Freely allows Titanium, may require as much as 100 kW, and with the
the flow of niobium cost of electric power now running from 10¢ to
electrons
20¢ per kilowatt per hour, it is obvious that
this portion of the operating expense can be
substantial.
flow, energy must be supplied. Energy is required
to move a charge, Q, through an electric poten-
MAGNETISM
tial, V. Power (P) is the rate at which energy (E)
is used or work (W) is performed. Like mass and charge, magnetism is a fundamen-
tal property of matter. All matter is magnetic to
Electric Power some degree. For example, an iron magnet picks
up a steel paper clip but not a copper penny. Steel
P = E/t = W/t ( joule/second ) is ferromagnetic and copper diamagnetic.
therefore Even subatomic particles such as protons have
P = QV/t = IV magnetic properties. Basically, magnetism is a
Alternatively, with application of Ohm’s law field effect associated with certain types of mate-
P = I 2 R = V 2 /R rial said to be magnetic. The magnetic field is
similar in many ways to the electric field, but its
The quantity for power, joule per second, is manifestation is different.
given a special name in physics, the watt (W). The earliest magnets were described 2000
Therefore the watt is the unit of electric power. years ago as naturally occurring black stones
that attracted iron. These “leading stones,” or
Energy and work are measured in joules. Power lodestones, were thought to be magic by the
is joules per second. natives in a region of present-day western Turkey,
then known as Magnesia in the Greek language.
In terms of human activity, the watt is very The term magnetism was adopted and persists
large. A construction worker may be able to today.
work sufficiently hard to power a few 100-W
light bulbs. In terms of human consumption, the
watt is a small quantity. In the United States,
Magnetic Domains
about 6 kilowatts (kW) per person is required The smallest region of magnetism is called a
continuously. In many of the developing nations, magnetic domain. Most materials have their
the figure is less than 500 W. magnetic domains randomly oriented (Figure
Resistive electromagnet-type MRI systems 2-9, A) and therefore exhibit no magnetism.
require considerable amounts of power. Some Some materials, however, have their magnetic
26 PART I  Fundamentals

TA B L E 2 - 2 The Three Magnetic States of Matter

Class Property Susceptibility Material

Ferromagnetism Easily magnetized >1 Iron, nickel, cobalt
Paramagnetism Very weakly magnetized 0-1 Aluminum, platinum, manganese, gadolinium
Diamagnetism Unaffected by a magnetic field <0 Copper, silver, mercury, carbon

domains aligned and therefore become magnets Paramagnetic materials include platinum,
(Figure 2-9, B). oxygen, tungsten, manganese, and gadolinium.
The strength and number of magnetic domains These materials have a magnetic susceptibility
in materials are associated with the materials’ less than 1. They are weakly influenced by an
electron configuration. As discussed later, an external magnetic field but do not exhibit mea-
electric charge in motion creates a magnetic field. surable magnetic properties of their own.
In the case of common magnetic materials and Diamagnetic materials have negative magnetic
electromagnetism, the magnetism is related to susceptibility and in fact are slightly repelled by
moving electrons. In the case of nuclei, the mag- magnets. Such materials include mercury, silver,
netism is related to the much weaker magnetic copper, carbon, and water.
properties of the spinning electrically positive
nucleus. Magnetic susceptibility relates the relative ease
The magnetic fields of paired electrons cancel; with which a material can be made magnetic.
therefore atoms with even numbers of electrons
in shells exhibit little magnetism. Atoms with
The Laws of Magnetism
unpaired electrons produce strong magnetic
domains. Atoms with nearly half-filled shells Dipoles.  Unlike the situation that exists with
have the strongest magnetism because electrons electricity, there is no smallest unit of magnetism.
generally will not begin pairing spins until a shell Because each magnetic domain exists with two
is half full. poles, a north pole and a south pole, it is com-
monly called a dipole. Unlike electric charge,
Electron configuration determines the three types a magnet cannot exist with a single pole. Divid-
of magnetism: ferromagnetism, paramagnetism, ing a magnet simply creates smaller magnets
and diamagnetism. (Figure 2-10).
Attraction/Repulsion.  As with electric charges,
Table 2-2 summarizes these types of magne- like magnetic poles repel and unlike magnetic
tism. Materials with such properties are distin- poles attract. In addition, by convention the
guished from each other according to the strength imaginary lines of the magnetic field leave the
and alignment of their magnetic domains in the north pole of a magnet (Figure 2-11) and return
presence of an external magnetic field. to the south pole. How do scientists know that
Ferromagnetic materials are easily magnetized these imaginary lines exist? They can be demon-
and have a magnetic susceptibility greater than strated by the action of iron filings near a magnet
1. Such materials include iron, cobalt, and nickel. (Figure 2-12).
These materials make the strongest magnets and The polar convention of magnetism actually
are used singly and in combination with MRI has its origin in the compass. The end of a
permanent magnets. Perhaps the most common compass needle that points to the Earth’s North
permanent magnet is one made of an alloy of Pole (actually, the Earth’s magnetic south pole)
aluminum, nickel and, cobalt—alnico. is the north pole of the compass. If a compass
 CHAPTER 2  Electricity and Magnetism 27

were taken to the North Pole, it would point into

the earth (Figure 2-13). At the South Pole, it
would point to the sky.
Magnetic Induction.  Just as an electrostatic
charge can be induced from one material to
another, nonmagnetic material can be made mag-
netic by induction. The magnetic field lines just
described are called magnetic lines of induction,
and the density of lines is proportional to the
intensity of the magnetic field.
When ferromagnetic material such as a piece
of soft iron is brought into the vicinity of an
intense magnetic field, the lines of induction
are altered by attraction to the soft iron (Figure
FIGURE 2-10  If a magnet is broken into smaller pieces, 2-14), and the iron will be made magnetic. If
baby magnets result.
copper, a diamagnetic material, were to replace
the soft iron, there would be no such effect.
This property of matter is called magnetic
permeability.
This principle is used with many MRI systems
N N S that use iron as a magnetic shield to reduce
the level of the fringe magnetic field. This is also
the basis of antimagnetic watches, although the
effect is not guaranteed near the strong field of
S
an MRI system.

FIGURE 2-11  The imaginary lines of the magnetic field Ferromagnetic material has high magnetic per-
leave the north pole and enter the south pole. meability and acts as a magnetic sink by drawing
the lines of induction into it.

When ferromagnetic material is removed

from the magnetic field, it usually does not retain
its strong magnetic property. Therefore soft iron
makes an excellent temporary magnet. It is a
magnet only while its magnetism is being induced.
If properly tempered by heat or exposed to an
external magnetic field for a long period, however,
ferromagnetic materials retain their magnetism
when removed from the external magnetic field
and become permanent magnets.
Another type of magnet is the electromagnet.
It owes its magnetism to an electric current,
which induces magnetism (Figure 2-15), but only
FIGURE 2-12  Demonstration of magnetic lines of force while the electric current is flowing. MRI systems
with iron filings. (Courtesy Robert Waggener, San Antonio, can use permanent magnets, resistive electromag-
TX.) nets, or superconductive electromagnets.
28 PART I  Fundamentals

Compass points
toward Earth

North
magnetic pole
Compass parallel with
Earth's surface

Compass points
skyward
FIGURE 2-13  A free-swinging compass reacts with the Earth as though it were a bar magnet.

The gravitational force is a long-range force.

Iron Copper
If it is assigned a relative value of 1, the electric
and magnetic forces have a value of 1037 times
N S N S
its magnitude.
The equation of interacting magnetic fields is
named for Karl Gauss, who used Coulomb’s law
to explain magnetism in the 1840s. The magnetic
FIGURE 2-14  Ferromagnetic material such as iron attracts
magnetic lines of induction, whereas nonmagnetic mate- force obeys the inverse square law principle, and
rial such as copper does not. its magnitude is proportional to the product of
the two interacting magnetic poles. Its formula-
tion is as follows:
Magnetic Force.  The force created by a mag-
netic field behaves similarly to that of the electric Gauss’s Law
field. The electric and magnetic forces were
F = k pP/d 2
joined by Maxwell’s field theory of electromag- where F = the force in newton (N)
netic radiation into a unified explanation. p and P = the relative pole strengths in ampere
Table 2-3 summarizes some of the similarities meter (Am)
of these three fundamental forces. The defining   d = the separation distance in meter (m)
equation is precisely the same among the three;   k = a constant
however, the magnitudes are different.
 CHAPTER 2  Electricity and Magnetism 29

FIGURE 2-15  Iron filings show the magnetic field lines of an electromagnet. (Courtesy Murray Solomon, San Jose, CA.)

TA B L E 2 - 3 Three Fundamental Forces

Gravitational Electric Magnetic

The force is: Attractive only Attractive and repulsive Attractive and repulsive
Acts in: Mass, m Charge, q Pole, p
Through an associated field: Gravitational field, g Electric field, E Magnetic field, B
With intensity: F = mg F = qE F = pB
The source of the field is: Mass, M Charge, Q Pole, P
The intensity of the field at a distance g = GM/d2 E = kQ/d2 B = kP/d2
from the source is:
The force between fields is given by: Newton’s law Coulomb’s law Gauss’s law
F = G Mm/d2 F = k Qq/d2 F = k Pp/d2
Where: G = 6.678 × 10−11 Nm2/C2 k = 9.0 × 109 Nm2/C2 k = 10−7W/A2

As with the gravitational force and the electric Magnetic Field

force, the magnetic force does not need a con-
ducting medium. It acts through space. B = F/p
where B is the field strength in tesla (T), F is the
force in newton (N), and p is the pole strength
The Magnetic Field in ampere meter (Am).
The imaginary lines of magnetic induction create
a field effect. The strength of the magnetic field The tesla (T) is the standard international (SI)
is defined by placing an imaginary north pole in unit for the magnetic field. An older unit still
it and measuring the force on the pole. This is very much in use is the gauss (G); 1 T equals
similar to the definition of the electric field by its 10,000 G. As with other types of fields, the
force acting on a positive charge. Therefore the strength of magnetic fields ranges over many
magnetic field B is given by the following: orders of magnitude.
30 PART I  Fundamentals

ELECTROMAGNETISM
N
A motionless electron has an electric field associ-
ated with it. An electron in motion has both an
S
electric and a magnetic field. The interaction
between the electric field and the magnetic field
is the basis for electromagnetism.

Oersted’s Experiment
Circuit open
Until the 1820s, electricity and magnetism
were considered two separate, unrelated, and
independent manifestations. As with many great
discoveries, Hans Christian Oersted accidentally N
noted that a compass was deflected by a DC
current. S
When no current exists in the circuit, the
compass needle placed close to the conductor
will point to the earth’s North Pole (Figure 2-16).
Once the switch is closed and current flows, the
compass immediately aligns itself perpendicu-
larly to the current-carrying wire. If the electron
flow is as illustrated in Figure 2-16, the north
pole of the compass is attracted to the wire as
shown. Reversing the current causes the south N
pole of the compass to point to the wire.
Oersted’s observation demonstrated that a
magnetic field is always associated with a moving S
charged particle. Furthermore, if either the elec-
tric field or magnetic field is time variant, that is,
changing in intensity with time, profound inter-
actions can occur.
The magnetic field induced by a moving elec-
tron is illustrated in Figure 2-17. The magnetic FIGURE 2-16  A compass is deflected by a direct current,
field lines extend radially from the axis of motion. and the direction of deflection depends on the direction
of the electric current.
How is it known that moving charged parti-
cles create a magnetic field? This cannot be
shown in an isolated frame of reference. Such a The electron moving in the magnetic field
demonstration requires another magnetic field so experiences a force that is at right angles to
that the interaction between the two magnetic both its velocity and the direction of the external
fields can be observed. magnetic field. This force tends to deviate the
An electron is shown moving out of this page electron in its motion, causing it to follow a
between the poles of a magnet (Figure 2-18); this curved path.
electron direction is indicated by the (Θ), which The direction of the force is given by the left-
represents the head of an arrow. If the electron hand rule (Figure 2-19). With application of the
were moving into the page, (X), representing the left-hand rule, it can be concluded that the elec-
tail of an arrow, would be shown. tron in Figure 2-18 should move down the page.
 CHAPTER 2  Electricity and Magnetism 31

Magnetic This is so because the external magnetic field and

field lines that of the electron reinforce one another above
the path of the electron and oppose one another
below that path. The magnitude of this force is
the following:

Magnetic Force
F = qvB
where F = the force in newton (N)
FIGURE 2-17  A moving charged particle induces a mag-   q = the charge in coulomb (C)
netic field in a plane perpendicular to its motion.
  v = the electron velocity in meter per
second (m/s)
  B = the magnetic field in tesla (T)
B fields add

This expression is yet another approach to

the definition of the magnetic field. A magnetic
N S field of 1 T exists when 1 C traveling at 1 m/s is
acted on by a force of 1 N (1 T = 1 N/Am).

One tesla is equal to one newton per ampere-

B fields oppose meter.
FIGURE 2-18  An electron moving in a magnetic field
experiences a force causing it to curve. The force results The total force on a moving electron is the sum
from the interaction between the electron’s magnetic field
and the external magnetic field. of the forces caused by external electric and mag-
netic fields.

Lorentzian Force
F = qE + qvB

This is called the Lorentzian force, and it is the

basis for such diverse yet readily recognizable
phenomena as the aurora borealis (northern
lights), the cathode-ray tube, and the operation
of a cyclotron for positron-emission tomography
Force (PET).
When electrons move in a conductor, the
B field Lorentzian force determines their use in electro-
mechanical devices such as motors and genera-
tors. In an MRI system, this force can be used to
explain the operation of the primary magnet, the
Velocity gradient coils, and the RF coils.
The Solenoid.  Still another method for defining
FIGURE 2-19  The left-hand rule demonstrates the direc- a magnetic field is based on the force on a long
tional relationships among force, velocity, and external section of current-carrying wire (Figure 2-20).
magnetic field. The external magnetic field, B, could be created
32 PART I  Fundamentals

1A
1N N S
F1N 1T
A.m

1A

1m
FIGURE 2-20  Parallel current-carrying wires repel each
other. This forms the basis for the definition of the tesla. FIGURE 2-21  A magnetic dipole is produced by a current-
carrying loop of wire.

by a permanent magnet or an adjacent current- Electric

carrying wire. The defining equation for the current
strength of the magnetic field is the following:
Magnetic
field
Magnetic Field
A
B = F/Idl
where F is force in newton (N), I is the current
Electric
in ampere (A), and dl is an incremental length current
(m) of a long wire.
Iron core More
intense
If the force on a 1-m section of wire conducting magnetic
field
1 A is 1 N, the magnetic field has a strength
of 1 T. B
FIGURE 2-22  A, A coil of current-carrying wire produces
a magnetic field. This is called a solenoid. B, With an iron
If the straight length of wire is shaped to form core, it is called an electromagnet.
a loop, a magnetic dipole is produced (Figure
2-21). If the current is reversed, the polarity of
the dipole is reversed.
If a series of loops is formed from a current- type of MRI magnet. It is also the foundation
carrying wire, a more intense magnetic field is for more complicated electromagnetic devices
produced (Figure 2-22, A). Such a helically such as motors, generators, and transformers. All
wound coil of wire is called a solenoid. these devices function because coils of current-
The Electromagnet.  If a rod of ferromagnetic carrying wire are wrapped around ferromagnetic
material is inserted inside the solenoid (Figure cores and magnetic fields are induced.
2-22, B), the intensity of the induced magnetic A motor is a device that converts electric
field is greatly strengthened because of the con- energy into mechanical energy. A generator does
centration of magnetic field lines. The device the opposite; it converts mechanical energy into
described is an electromagnet, and it is the basis electric energy. A transformer alters the magni-
for switches, large industrial magnets, and one tude of electric current and voltage.
 CHAPTER 2  Electricity and Magnetism 33

No electric current Electric current Reverse electric current

A B C
FIGURE 2-23  A, No electric current exists in a closed circuit that has no source of electric potential. B, When a magnetic
field moves through a closed coil of wire, an electric potential is created and an electric current induced. C, Reverse the
magnet, and the electric current reverses.

The intermediate step in each of these devices Faraday’s Law

is the induced magnetic field. Mechanical motion
can be extracted from electric power or supplied V = -dB/dt
to produce electric power because of the force on where V is the induced voltage (V), dB represents
a current-carrying wire in the presence of a mag- the changing magnetic field, and dt is the time
netic field. taken for that change.

The negative sign in Faraday’s law is a conse-

Faraday’s Law quence of Lenz’s law. The direction of the induced
A force is exerted on a length of current-carrying electron flow, and therefore the induced voltage,
wire when a magnetic field is present. The force is such that it opposes the agent inducing the
is perpendicular to both the magnetic field and flow.
the direction of electron flow. The electron flow induced by moving the
If a loop of wire is connected to an ammeter north pole of the permanent magnet toward the
to produce a closed circuit with no source of loop is in such a direction that the north pole of
electric potential, such as a battery, there is no the induced magnetic field opposes a further
electric current in the loop (Figure 2-23, A). If a push of the magnet (see Figure 2-23, B). If the
permanent magnet is moved through the loop, permanent magnet is pulled away from the loop
as in Figure 2-23, B, a current will flow. If (see Figure 2-23, C) or the south pole of the
the movement of the permanent magnet is magnet is pushed toward the loop, the induced
reversed, the electric current is reversed (Figure electron flow is reversed, and the polarity of the
2-23, C). induced magnetic field is reversed. Regardless of
The changing magnetic field created by the whether the magnetic field or the loop of wire is
moving permanent magnet induces a voltage in moved, a current is induced, which in turn
the circuit, causing electrons to flow. This phe- induces a secondary magnetic field to oppose the
nomenon was demonstrated by Michael Faraday inducing field.
in the 1830s and is stated as Faraday’s law of If an opposing loop of wire is used instead of
electromagnetic induction. a permanent magnet as the inducing agent, a
34 PART I  Fundamentals

simple transformer is made (Figure 2-24). If the The Transformer.  This principle is the basis for
first coil is energized by a source of AC power, the transformer. A transformer incorporates hun-
the magnetic field associated with the first coil dreds of loops of wire coiled on a ferromagnetic
alternates in polarity. This primary alternating core. The core concentrates and intensifies the
magnetic field interacts with the second coil as magnetic field. Because a moving magnetic field
though one alternately pushed and then pulled a is required, a transformer will not work on DC,
permanent magnet along the axis of the second only on AC.
coil. The induced electron flow in the secondary The RF Coil.  The current in a primary coil
coil is AC, and the induced magnetic field always induces a current in a secondary coil through
opposes the action of the primary field. magnetic induction, but only when the primary
current varies in intensity (see Figure 2-24).
However, the phenomenon of Faraday induction
can be carried one step further. If electrons are
not only varying in intensity but also accelerating
and decelerating alternately, electromagnetic
radiation is emitted.
This is the principle of the radio, in which case
the electromagnetic radiation is called RF (Figure
2-25). Oscillating electrons in the transmitting
antenna (coil) produce RF, which in turn induces
a signal in the receiving antenna (coil). This is
similar to the way that the net magnetization in
the patient interacts with the RF coils in MRI.

Secondary ELECTROMAGNETIC RADIATION

coil
Primary As previously discussed, a resting electric charge
coil radiates an electric field. When the charge is in
FIGURE 2-24  An alternating current in one coil creates motion, a magnetic field is generated.
an alternating magnetic field that can induce an alternat- When the moving electric charge slows
ing current in a nearby second coil. down (decelerates), a photon of electromagnetic

FIGURE 2-25  Magnetic resonance imaging has several characteristics of a radio.

 CHAPTER 2  Electricity and Magnetism 35

radiation is emitted. Radiologists and radiogra- Consequently, ultrasound is known as a longitu-

phers know that bremsstrahlung x-rays are pro- dinal wave. The terms transverse and longitudi-
duced by the deceleration of a projectile electron nal are also used to describe the two MRI
in the vicinity of the nucleus of an atom in the relaxation times (T2 and T1, respectively);
anode of an x-ray tube. however, when so used in MRI, they are not
Electrons decelerated or accelerated within related to waves.
the conducting element of a radio coil emit The intensity of each photon begins at zero,
photons of RF. This phenomenon was described increases rapidly to a maximum, and decreases
mathematically in the 1860s by James Clerk to zero again. Consequently, each photon does
Maxwell. have an origin and a termination. The energy of
the electromagnetic photon is determined by the
amount of kinetic energy lost by the charged
Maxwell’s Wave Equations particle.
Maxwell synthesized the then-known laws of Maxwell’s equations also demonstrated that
electricity and magnetism into what are now electromagnetic photons obey the classical wave
known as Maxwell’s wave equations of electro- equation.
magnetic fields. At that time, the only electro-
magnetic radiation recognized was visible light. Wave Equation
Maxwell described light mathematically in
terms of oscillating electric and magnetic fields v = fl
and showed that the interaction of these fields where λ is the photon wavelength (m), ƒ is the
caused a wave to be propagated in free space at photon frequency (Hz), and v is the velocity
a direction 90° to the fields. Furthermore, the (m/s).
velocity of propagation, c, was later shown to be
3 × 108 m/s.
The Electromagnetic Spectrum
Maxwell’s mathematical equations describe a
photon of electromagnetic radiation (Figure Mention of Albert Einstein’s 1905 theory of
2-26). The electric and magnetic fields are at relativity and Max Planck’s 1915 formulation
right angles to one another and are also at right of quantum mechanics is required to complete
angles to the velocity vector, c. a discussion of electromagnetic radiation.
When the field disturbance is perpendicular to Einstein showed that electromagnetic photons
the direction of propagation, such radiation is behaved relativistically; that is, if two electrons
known as a transverse wave, in contrast with are decelerated together, stationary observers
an ultrasound wave, where tissue molecules would observe the emission of two photons.
oscillate in the direction of the velocity vector. However, if those same observers were on one of

E E
C

B
B
C

FIGURE 2-26  An electromagnetic photon consists of electric and magnetic fields oscillating at right angles to one
another and traveling at the speed of light.
36 PART I  Fundamentals

the electrons, they would be able to verify that interact as a particle or a wave, depending on the
the neighboring electron had only an electric field photon energy.
and no electromagnetic radiation would be Electromagnetic radiation extends over an
detected. enormously wide range (see Figure 1-1). This
Furthermore, Einstein showed not only that span of electromagnetic radiation is known
energy was conserved by the conversion of kinetic as the electromagnetic spectrum, and it extends
energy into electromagnetic energy but also that from radio emissions on the low-energy side
electromagnetic energy could represent the con- through microwaves, infrared, visible light,
version of energy into matter, according to his and ultraviolet to high-energy x-radiation and
famous equation. gamma-radiation.
This representation of the electromagnetic
Relativity spectrum contains three scales, E, λ, and ƒ, each
of which is equivalent according to Planck’s
E = mc 2 quantum theory. In addition, each scale has some
where E is the energy in joule (J), m is the particle historical interest, because it reflects the manner
mass (kg), and c is the photon velocity (3 × in which the photons are produced and interact
108 m/s). with matter.
Light was the earliest electromagnetic radia-
By the time of Max Planck, the various forms tion to be studied. It was shown to interact pri-
of electromagnetic energy were recognized as marily as a wave and therefore was usually
being different manifestations of a similar funda- characterized by its wavelength.
mental disturbance. The great body of physical Radio emissions are produced by oscillating
measurements resulting from the previous 100 electrons energized by a special electric circuit
years’ experience with light and optics held that called an oscillator. This was first demonstrated
the interaction of electromagnetic radiation with by Heinrich Hertz in the 1880s. Because the
matter was wavelike. oscillation of emission is the principal design
Planck’s quantum theory described electro- parameter, these waves are identified by their
magnetic radiation as consisting of particles, frequency.
called photons, and showed the way such radia- Roentgen discovered x-rays in 1895, and they
tion could behave as a particle during its inter­ were characterized by the voltage of production.
action with matter. The proof of this is the When Einstein and Planck explained x-ray inter-
photoelectric effect described by Einstein and action with matter as particles, it became conven-
the fundamental equation underlying Planck’s tion to identify such photons by their energy.
theory. Interest in diagnostic imaging lies in these
three distinct regions of the electromagnetic
Plank’s Quantum Theory spectrum called imaging windows. Each of the
three windows is described by one of the three
E = hf = hc/ l scales.
where E is the energy of the photon (J); h
is Planck’s constant, 6.63 × 10-34 Js; ƒ is fre- An imaging window is a range within the elec-
quency; c is the velocity of light; and λ is the tromagnetic spectrum used to produce images.
wavelength (m).
The Visible Window.  Visible light interacts
Planck further showed that electromagnetic with matter more like a wave than a particle.
radiation indeed actually possessed a duality of Diffraction, refraction, reflection, and interfer-
nature insofar as its interaction with matter ence are all properties of wavelike interactions,
was concerned. Electromagnetic radiation does and they all apply to visible light.
 CHAPTER 2  Electricity and Magnetism 37

Visible light photons’ wavelengths range from

approximately 400 to 700 nm. Visible light is
produced essentially by electron shifts from an
excited energy state to the ground state of the
outermost shell of an atom. Its wavelength allows
it to interact with the receptor cells of the retina,
but its quantum energy is too low to ionize
matter.
Imaging with visible light occurs by sensing
the reflection of light from a patient. Therefore
the image of the patient is a surface image and
not an image of interior structures. Individual
light photons are detected by specially evolved
optical light sensors of the human eye, the rods
and cones, or by the special molecules in the
emulsion of a photographic film. FIGURE 2-27  X-ray images are similar to a shadowgram
The X-ray Window.  X-rays are produced by except that they provide outlines of internal structures as
changes in the kinetic energy of fast-moving well as the surface. (Dedicated to Xie Nan Zhu, Guang-
electrons. Bremsstrahlung x-rays can be viewed zhou, People’s Republic of China.)
simplistically as being created by the deceleration
of a high-velocity electron in the vicinity of the RF is used extensively in communications
nucleus of a target atom. Characteristic x-rays (television, radio, and microwave). Standard
are produced when the outer-shell electrons commercial AM broadcast operates from ap-
are decelerated at an inner shell. The energy proximately 540 to 1640 kHz. FM radio and
of photons used for x-ray imaging ranges television occupy a band of frequencies from
from approximately 20 to 150 kiloelectron approximately 50 to 200 MHz. This range over-
volts (keV). laps with the range used in MRI.
X-rays are used to image the body in much Many sources of RF radiation exist in this
the same way as a shadowgram is produced with frequency range, all of which can interfere with
light (Figure 2-27). Of course, in a shadowgram, the MRI signal. Measures usually must be taken
only the outline of the figure is imaged. During to ensure that these extraneous sources of RF are
x-ray examination, not only the outline but also attenuated or entirely excluded from the coil
the internal structures are imaged. used to receive the MRI signals.
X-ray imaging is possible because of the This presents a problem similar to that encoun-
particle-like interactions between x-rays and tered in x-ray imaging. An x-ray examination
tissue atoms. These interactions occur principally room is shielded to ensure that x-radiation does
by way of photoelectric effect and Compton scat- not escape from the room and create a radiation
tering, although at low energies coherent scatter- hazard to persons nearby. An MRI room may
ing may also contribute to the attenuation of the require shielding to exclude extraneous RF from
x-ray beam. Therefore a radiograph represents the imaging system. Such a shielded room, called
the pattern of x-ray attenuation while passing a Faraday cage, is discussed in Chapter 13.
through the patient. The resulting image is a
function of the x-ray attenuation coefficient.
CHALLENGE QUESTIONS
The MRI Window.  Electromagnetic radia-
tion with frequencies of approximately 10 to 1. In an MRI system, what produces the static
200 MHz is used in MRI. This radiation is in the magnetic field (B0) and the gradient magnetic
RF portion of the electromagnetic spectrum. fields (BXYZ)?
38 PART I  Fundamentals

2. X-ray imaging rooms must be shielded 7. The electric field is force exerted per electro-
(usually with lead) to reduce the radiation static charge. What are the units of the
exposure of persons outside of the room. electric field, and what does each unit
Why are MRI rooms shielded? represent?
3. Who is considered the father of the study 8. The universal wave equation applies to all
of electrostatics, and precisely what is harmonic motion and in particular to diag-
electrostatics? nostic ultrasound, x-ray imaging, and MRI.
4. X-rays resemble RF radiation used in MRI State the wave equation and the units for
in many ways. What are the similarities and each parameter.
differences? 9. State the four principal laws of electrostatics.
5. The physical basis for MRI is electromag- 10. How is a magnetic field defined and
netic induction. Just what does the word measured?
induction mean in this sense?
6. What is the underlying equation that
describes quantum mechanics?
 CHAPTER

3 
Nuclear Magnetism

OBJECTIVES
At the completion of this chapter, the student • Describe the appearance of net magnetization
should be able to do the following: on a vector diagram.
• Distinguish between classical mechanics and • Discuss the factors that influence the
quantum mechanics. magnitude of net magnetization.
• State the Larmor equation and relate its • Distinguish between stationary frame and
importance to magnetic resonance imaging rotating frame and state which is used in
(MRI). MRI.

OUTLINE
Quantum Mechanical Control of Net
Description Magnetization
Classical Mechanical Reference Frames
Description Stationary Frame
Net Magnetization Rotating Frame
Vector Diagrams

KEY TERMS
Classical mechanics Gyromagnetic ratio Quantum mechanics
Equilibrium Net magnetization

As the name nuclear magnetic resonance (NMR) of the twentieth century and is the basis for con-
indicates, the magnetic resonance imaging (MRI) temporary investigations into the structure of
signal originates from the nuclei of atoms reso- matter. Quantum mechanics, though, is complex
nating in a patient in the presence of a magnetic and does not provide an intuitive understanding
field. Because this is a nuclear phenomenon, an of what is really happening during an MRI
accurate and complete description requires the examination at the subatomic level.
use of quantum mechanics. Fortunately, many of the concepts that relate
Quantum mechanics is the branch of physics to how MR images are formed relate to net
that describes the behavior of very small objects, magnetizations that represent the summed prop-
such as x-rays, protons, neutrons, and electrons. erties of billions and billions of nuclear magnets,
This branch of physics evolved in the first quarter say in a drop of water. Therefore, we can think

39
40 PART I  Fundamentals

certain precisely allowed states of spin. The

allowed values are 0, 1 2 , 1, 3 2 , and so on.
Each nucleus has its own characteristic spin
quantum number. For example, all hydrogen
atoms of atomic mass 1 have a spin quantum
number of 1 2 . All carbon atoms of atomic mass
12 have a spin quantum number of 0. All carbon
atoms of atomic mass 13 have a spin quantum
Classical Quantum
number of 1 2 .
mechanics mechanics
The spin quantum number dictates many of
FIGURE 3-1  The motion of large objects is described by
classical mechanics, that of subatomic particles and the magnetic resonance (MR) properties of a
photons by quantum mechanics. given nuclear species. Table 3-1 presents some
nuclei of interest to MRI and their respective spin
quantum numbers. Once this quantum mechani-
of the nuclear magnetization, as it is manifested cal premise is accepted, it is easy to use classical
in the cellular water of a patient, as being mechanics to describe the way the MR signal is
largely described by the principles of classical generated.
mechanics. One result of the spin being quantized is that
there are a limited number of ways a nucleus can
Classical mechanics deals with large objects; spin. Each of these ways is called a spin state.
quantum mechanics deals with subatomic parti- For a spin 1 2 nucleus, there are only two allowed
cles and photons. spin states, + 1 2 and − 1 2 .
In general, for a particle with spin S, there are
Classical mechanics is the branch of physics 2S possible spin states, and each spin state is one
that describes the behavior of large objects such integer value separated from the next. Thus for
as rockets, automobiles, and Ping-Pong balls. It a nucleus such as 23Na with a spin quantum
has its origins in the seventeenth-century ideas of number of 3 2 , the allowed spin states are + 3 2 ,
Sir Isaac Newton. + 1 2 , − 1 2 , and − 3 2 .
The differences between these two branches A fundamental law of physics states that a
of physics are demonstrated in Figure 3-1. spinning, charged mass induces a magnetic field
For many cases, the statistical averaging that about itself. Earth is one such example. The
occurs over a patient reduces the quantum nucleus is a spinning, charged particle and there-
mechanical description to a classical mechanical fore has an associated magnetic field. This field
description. One quantum mechanical concept, is given the special name of nuclear magnetic
however, that cannot be overlooked or avoided moment, and its intensity is related to the mass,
is nuclear spin. charge, and rate of spin of the nucleus.

A spinning proton induces nuclear magnetism.

QUANTUM MECHANICAL
DESCRIPTION
According to quantum mechanics, every nucleus Thus each spin state has a different nuclear
has a quantity called spin. This quantity is not magnetic moment. In the absence of an external
exactly what is normally thought of as spin, but magnetic field, each spin state has the same
it is close enough. energy, but in the presence of an external mag-
There is one anomaly, though; this spin is netic field, each spin state has a different energy
quantized into units of half-integer values and and this energy difference increases as the exter-
called the spin quantum number. There are only nal magnetic field increases.
 CHAPTER 3  Nuclear Magnetism 41

TA B L E 3 - 1 Magnetic Resonance Imaging Properties of Medically Important Nuclei

Isotope Spin Quantum Number Gyromagnetic Ratio (MHz/T) % Abundance

1
H 1
2 42.6 99
12
C 0 0 98
13
C 1
2 10.7 1.1
16
O 0 0 99
17
O 5
2 5.8 0.1
19
F 1
2 40.0 100
23
Na 3
2 11.3 100
25
Mg 5
2 2.6 10
31
P 1
2
17.2 100
33
S 3
2
3.3 0.7
57
Fe 1
2 1.4 2.2

A nucleus in the presence of an external mag-

netic field prefers to be in the lower energy state
rather than in the higher energy state. Therefore
the lower energy state has more nuclei than the N N
higher energy state. Such a system of nuclear RF
spins is said to be at equilibrium with the exter- B0
nal magnetic field.
With the use of electromagnetic radiation that S S
has an energy exactly equal to the energy differ-
ence between two nuclear energy states, this
population difference can be disturbed by causing
some nuclei of a lower energy state to absorb Equilibrium Energized
energy and join the higher energy nuclei (Figure FIGURE 3-2  At equilibrium, more proton spins are
3-2). The system of energized nuclear spins aligned with the magnetic field and said to be in the low-
energy state. When energized, more proton spins are
slowly returns to its original population differ- aligned against the field.
ence while emitting a signal that can be observed.
This is the MRI signal.

CLASSICAL MECHANICAL
DESCRIPTION
From the classical mechanics point of view, the N
nucleus is simply a charged particle that is spin-
ning. The spinning motion generates a magnetic
field parallel to the axis of spin. P
This magnetic field is given the same name of µ
nuclear magnetic moment (µ) that was described
in quantum mechanics (Figure 3-3). It is referred S
to with the Greek letter mu (µ). For any indi- FIGURE 3-3  A representation of the nuclear magnetic
vidual nucleus, the orientation of the axis of moment (µ) of a hydrogen nucleus.
42 PART I  Fundamentals

S
B0

N
S
S

NS
N

FIGURE 3-4  The behavior of horseshoe magnets is to

align parallel with opposing polarity.
P µ

FIGURE 3-6  Precession of a nuclear magnetic moment

in the presence of an external magnetic field.

Similarly, when a nuclear magnetic moment is

Gravity placed in the presence of an intense static mag-
netic field, B0, its axis of rotation precesses about
the magnetic field (Figure 3-6). The exact fre-
quency of precession by such a nucleus can be
calculated with the Larmor equation.

FIGURE 3-5  The motion of a spinning gyroscope in the Larmor Equation

presence of gravity.
f = gB0
where ƒ is the frequency of precession (MHz), γ
rotation and therefore the direction of the nuclear is the gyromagnetic ratio (MHz/T), and B0 is the
magnetic moment is random. strength of the external magnetic field (T).
For MRI, the question is how this nuclear
magnetic moment will interact with an externally This equation demonstrates that there is a
applied magnetic field (B0). Classically, when two strictly linear relationship between the frequency
magnets are near, a force is generated that tends of precession, f, and the strength of the external
to make the two magnetic fields become parallel magnetic field, B0. The gyromagnetic ratio is
and pointed in the opposite direction. For horse- unique for each type of nucleus and must be
shoe magnets, this force is manifested by the measured for each. Thus it is an empirically
magnets’ aligning in opposition to each other determined factor used to convert field strength
(Figure 3-4). to precessional frequency. The values vary widely
However, the nucleus is a spinning particle, and can be found in Table 3-1 for several nuclei
and thus its actual movement in response to an of interest to MRI. Note that nuclei that have a
applied force such as the external magnetic field, gyromagnetic ratio = 0 will not be able to produce
B0, is the same as that produced by a gyroscope; magnetic resonance.
it moves perpendicular to the applied force This important relationship is used to explain
(Figure 3-5). In the case of the gyroscope, the many phenomena for both MRI and NMR
interaction is mass-mass: the mass of the Earth spectroscopy. The key factor is not what the
and the mass of the gyroscope. strength of the external magnetic field is but
 CHAPTER 3  Nuclear Magnetism 43

what magnetic field is experienced by the nucleus the behavior of each of the individual nuclei, and
as modified by many other environmental therefore the net magnetization (M) can be used.
influences. M is simply the sum of the individual nuclear
For spectroscopy this includes the contribu- magnetic moments, µ.
tion from the magnetic fields generated by the
motion of the orbiting electrons, which results Net Magnetization
in chemical shifts. It also includes time-varying
magnetic fields from nearby moving, polarized M = Sm
molecules that contribute to T2 relaxation and
magnetic field gradients at the interface of two With this equation, the properties of net mag-
tissues with different susceptibilities, which result netization during various circumstances can be
in T2* relaxation. Each of these parameters is considered. Because M is just the sum of many
dealt with later (see Chapter 7). µ’s, the general behavior of M is identical to the
For high-resolution NMR spectroscopy, mag- behavior of the individual µ’s.
netic field strengths are often described in terms Consider the armadillo shown in Figure 3-7
of the Larmor frequency at which hydrogen crossing a highway outside of College Station,
atoms resonate. Thus a 2.35-T magnet is often Texas, where the Earth’s magnetic field is negli-
referred to as a 100-MHz magnet, and an 11.7-T gible. The orientation of the spin axis of each
magnet is a 500-MHz magnet. Such odd- nucleus is random. Therefore the orientation of
numbered field strengths are often used to get the an individual nuclear magnetic moment pointing
resonant frequency of hydrogen atoms to mul- in any given direction is canceled by another
tiples of 50 or 100 MHz. nucleus, with its magnetic moment pointing
For the imaging world, this convenience has exactly opposite to the first. Consequently, the
not yet been achieved. Field strength values such sum of a large number of nuclear magnetic
as 0.5, 1.0, 1.5, 3.0, and 7.0 T are common, moments averages out to zero, or M = 0. This
although they result in approximate hydrogen agrees with the everyday observation that arma-
resonant frequencies of 21, 42, 63, 126, and dillos are totally nonmagnetic.
298 MHz, respectively.
Vector Diagrams
Net Magnetization Also shown in Figure 3-7 are three axes (X, Y,
Because measuring the action of individual nuclei and Z) drawn perpendicular to one another to
is impossible, any signals received or data col- describe a coordinate system. The illustration
lected during the MRI process are the result of a shows the Cartesian coordinate system used by
bulk phenomenon from perhaps as many as 1026 mathematicians to diagram phenomena in space.
nuclei. This is the approximate number of nuclei For MRI, such a coordinate system is used to
in a patient. construct vector diagrams.
A vector is a quantity that has direction. If a
Individual nuclei precess in the presence of an speed of 50 mph is discussed, that is a scalar
external magnetic field. quantity. If traveling 50 mph in a northerly direc-
tion is discussed, that is a vector quantity. The
A single, isolated nucleus is never observed, net magnetization, M, in Figure 3-7 is a vector
just collections of similar nuclei as an aggregate. quantity with a magnitude of zero. Vector dia-
This aggregate of spins is sometimes called an grams are extremely helpful in describing MRI
ensemble of spins to emphasize the bulk of the phenomena.
signal-producing nuclear spins. Fortunately, the When the armadillo is brought to the labora-
signal from a spin ensemble accurately reflects tory and placed in an external magnetic field,
44 PART I  Fundamentals

Y
M
X

FIGURE 3-7  The net magnetization of an armadillo in the absence of an external magnetic field is zero.

molecular nature of the tissues of the armadillo.

Some tissues align quickly and others more
slowly. One result of this alignment is that the
net magnetization will not be zero.
Z M The external magnetic field provides a
small preference for the spins to align with
the field. This is a small force compared with
B0
Y normal thermal interactions, and only a small
number of spins align with the external magnetic
X
field. The result can be illustrated as the vector
in Figure 3-8, which shows that net magnetiza-
tion exists along the Z-axis, parallel to the exter-
nal field.
If the armadillo is left in the magnet for a suf-
FIGURE 3-8  The net magnetization of an armadillo ficiently long time, the number of spins oriented
resting in a laboratory magnet has magnitude and with the field stabilizes to an equilibrium value.
direction. This value is referred to as M0 or the net magne-
tization at equilibrium, and it is the largest pos-
sible value of M.
something different happens (Figure 3-8). Clas- By convention, the external magnetic field is
sically, all the nuclei should precess about the parallel to the Z-axis of the Cartesian coordinate
applied field, and the net magnet moment should system; therefore this magnetization can also be
remain zero. referred to as MZ or the Z component of the net
Quantum mechanics predicts that more nuclei magnetization. At equilibrium the X and Y com-
will be in one state rather than in another. This ponents of net magnetization are zero, so MX =
nuclear alignment does not occur instantly but MY = 0 because all of the equilibrium magnetiza-
rather over a period of time determined by the tion is along the Z-axis.
 CHAPTER 3  Nuclear Magnetism 45

At equilibrium, MZ = M0; MX = MY = 0. As the temperature (T) of tissue decreases,

the net magnetization also increases. However,
because this is temperature in Kelvin (K), drastic
The intensity of the MR signal is related to the cooling must be used to obtain any significant
value of M0. The larger the M0, the stronger the increase in the MR signal.
MR signal. The number of nuclei contributing to
M0 is small (only about one of every 1 million
Control of Net Magnetization
nuclei contribute to M0), and MRI is therefore a
technique that has a poor signal-to-noise ratio. The net magnetization along the Z-axis, MZ, is
In addition to the number of contributing nuclei, invisible. It can be thought of as being stored
other factors affect the amount of signal avail- energy, and can be used in a similar way that we
able for MRI. These factors can be summarized use the electrical energy stored in a battery. From
by the following: the quantum mechanical point of view, directly
measuring this magnetization is impossible; from
the classical point of view, this magnetization,
Net Magnetization at Equilibrium MZ, is so small relative to B0 that it cannot be
PDg 2 B0 measured but needs to be converted to a form
M0 ∝ where it can be distinguished from B0. This fact
T
will become more apparent later. Essentially, it is
where α = symbol meaning proportional to
not possible to measure MZ; only MX and MY,
  PD = the number of nuclei (proton density)
which are collectively represented by MXY, can be
in the volume of interest
measured.
  T = the temperature of the material under
investigation
Mxy is net magnetization in the transverse
  γ = the gyromagnetic ratio
plane (XY plane) and is called transverse
  B0 = the strength of the external magnetic
magnetization.
field
For MZ to be observed, the net magnetization
Thus as the concentration of the nuclear species must be rotated (or flipped) off the Z-axis and
or the volume of tissue being sampled increases, into the XY plane. M, the net magnetization,
the MR signal increases because the proton is the sum of many nuclear magnetic moments
density (PD) is greater. and behaves in the same way that the individual
Hydrogen has the highest γ of any nucleus nuclear magnetic moments behave. Therefore,
other than tritium (3H) and has the second largest if the net magnetization is pointing anywhere
signal per number of nuclei. Tritium would there- except exactly along the Z-axis, it precesses
fore make an excellent MRI agent, but patients about the Z-axis with the same frequency at
would frown on its use because it is radioactive. which the individual nuclear magnetic moments
For 13C, γ is about one fourth the value for 1H, precess, the Larmor frequency.
and for an equal number of nuclei, the signal Precession is a form of change of the magnetic
available is much less. field, and this changing magnetic field induces an
At higher magnetic field strength (B0), the electric current in a loop of wire or coil placed
increase in net magnetization results in a stronger nearby. The signal can be detected, measured,
MR signal and therefore improvement in any and recorded.
MR image. This is the major reason for going to How is the net magnetization flipped off the
higher magnetic fields for human imaging systems Z-axis, and what precisely does this mean? Flip-
and spectroscopy. It is only one factor, albeit a ping the net magnetization describes the direc-
very important one. tion in which the net magnetization of the patient
46 PART I  Fundamentals

is altered. Instead of pointing straight up, it will One convention within this system is that the
now point somewhere else, actually anywhere applied magnetic field is always parallel to the
else. It is just as big as it was before but has a Z-axis. Although the Z-axis is always drawn as
new direction. up, its actual orientation is determined by the
direction of B0 in the particular magnet system
(Figure 3-10).
REFERENCE FRAMES The B0 field for a permanent magnet is vertical
The concept of a frame of reference is necessary and therefore passes through the supine patient
to follow the motions of the individual nuclear in the anterior-posterior direction. For some
magnetic moments and the net magnetization resistive and most superconducting electromag-
vector. The frame of reference used is the stan- nets, however, the B0 is horizontal and parallel
dard three-dimensional Cartesian coordinate with the axis of the body and therefore passes
system shown in Figure 3-9. through the patient from head to toe.

Z Stationary Frame
The view of someone standing next to the magnet
is referred to as the stationary frame of reference
because all motions are compared with someone
B0 standing still in the imaging room.
Y
A second magnetic field with special proper-
ties must be used to rotate (or flip) the net mag-
X netization away from the Z-axis. The second
FIGURE 3-9  The conventional three-dimensional coordi- magnetic field must precess about the B0 mag-
nate system used to describe the motion of the net mag- netic field with the same frequency as the nuclei,
netization vector. which is the Larmor frequency.

B0 B0

Z Z Z

Y Y B0 Y

X X X
Resistive Permanent Superconductive
FIGURE 3-10  The orientation of B0 and the stationary frame of reference in three types of magnetic resonance imaging
systems.
 CHAPTER 3  Nuclear Magnetism 47

If it does not precess at the Larmor frequency,

no interaction between the second magnetic
field and the nuclear spins is produced. This
phenomenon is called resonance because of the
requirement that the second magnetic field pre-
cesses at exactly the correct frequency or else
nothing will happen to the net magnetization
vector.

The second magnetic field must precess with the

frequency of the precessing nuclear magnetic
moments.

Locating such a magnetic field may seem dif-

ficult, but it is not. An electromagnetic emission
such as a radiofrequency (RF) is composed of an
oscillating electric field positioned 90° to an
oscillating magnetic field (see Figure 2-26).
The magnetic field component of an RF emis-
sion at the Larmor frequency is effectively a mag-
netic field rotating at the Larmor frequency. The
precessing nuclear magnetic moments interact
with the magnetic field component of the RF. FIGURE 3-11  The carousel serves to explain the rotating
Thus through irradiation of the sample with an frame. The only way to carry on a conversation with a
RF emission, a rotating magnetic field is pro- friend is to jump on the carousel and rotate too.
duced to flip the net magnetization vector into
the XY plane.
person, who can now converse easily. The rest
of the world is rotating “backwards” relative
Rotating Frame to them.
Because the net magnetization is precessing so The solution is to use a frame of reference that
fast, visualizing its motion is difficult. A trick is exactly matches the motion of the net magnetiza-
used to make it easy to follow the motion of the tion. This is a reference frame precessing about
net magnetization. An example of this can be the Z-axis of the stationary frame of reference
found in an amusement park. and with the same Z-axis. This new frame of
Someone in line to go on a carousel and a reference is called the rotating frame.
friend already on the carousel face the same The stationary frame of reference has axes
problem presented by the precession of the net labeled X, Y, and Z; the rotating frame uses axes
magnetization in the stationary frame of refer- X′, Y′, and Z′ to avoid confusion. Unfortunately,
ence (Figure 3-11). The rotation of the friend on because essentially all MRI vector diagrams use
the carousel relative to the person on the ground the rotating frame of reference, most authors
makes it difficult to carry on a conversation. The leave the primes off the axes for the rotating
friend may even be sitting on a horse that is going frame. This convention is followed in the remain-
up and down while the carousel spins, further der of this book.
complicating the motion. Within the rotating frame of reference, the net
However, if the person steps onto the magnetization does not precess and is thus easier
carousel, the friend is stationary relative to the to follow. The motion of the net magnetization
48 PART I  Fundamentals

Z Each imaging system has its own combination

of RF time and intensity; this is related to the RF
power and pulse duration used. Another interest-
ing stopping point for the flipped net magnetiza-
Y
tion is −Z or rotation of the net magnetization
through an arc of 180°. The combination of RF
X power and duration of the pulse that does this is
called a 180° RF pulse.

BX
FIGURE 3-12  In the rotating frame, a secondary mag- CHALLENGE QUESTIONS
netic field (BX) from a radiofrequency emission causes net 1. Discuss the differences between the two
magnetization to rotate in the YZ plane.
principal areas of physics, classical mechan-
ics and quantum mechanics.
2. What frame of reference is used when
describing MRI vector diagrams?
when the second magnetic field is turned on can 3. What is a magnetic moment?
now be followed. 4. If hydrogen nuclei, protons, precess at a fre-
In the rotating frame of reference, the second quency of 42 MHz, where could one possi-
magnetic field, BX, is stationary and is usually bly find another alternating field of the same
aligned along the X-axis (Figure 3-12). It can, frequency?
however, be aligned along the Y-axis just as 5. When an ensemble of proton spins is placed
easily. in an external magnetic field, some align
This new magnetic field causes the net magne- with the external field and others align
tization to precess around in the YZ plane with against. Which has more aligned spins, and
a frequency of γBX. BX can be turned off when which spins are in a higher energy state?
the net magnetization has reached any given 6. How are longitudinal magnetization and
location in the YZ plane. One such point is the transverse magnetization symbolized?
XY plane. At this point, all of the net magnetiza- 7. What is meant by the term equilibrium?
tion has been moved into the horizontal plane 8. Precisely, what determines the value of net
and will generate the maximum MR signal magnetization at equilibrium, and how is
possible. this value symbolized?
The conditions of strength of BX used to 9. Nuclear magnetic spectroscopy is usually
accomplish this are called a 90° RF pulse because identified by the frequency of analysis, not
the net magnetization rotates through an arc of the magnetic field intensity. Why?
90°. These conditions are determined by the time 10. Relate the equation that describes net mag-
and intensity of the RF pulse. netization and identify each parameter.
 CHAPTER

4 
Equilibrium–Saturation

OBJECTIVES
At the completion of this chapter, the student • Identify the primary magnetic resonance (MR)
should be able to do the following: signal.
• Define net magnetization and its relationship • Draw a two-line RF pulse sequence and label
to equilibrium, partial saturation, and all of its components.
saturation.
• Distinguish between a hard and a soft
radiofrequency (RF) pulse.

OUTLINE
Net Magnetization XY Magnetization Free Induction Decay
Equilibrium MXY Precession Radiofrequency Pulse
Flip Angle Dephasing of XY Diagrams
Radiofrequency Pulses Magnetization
Hard and Soft Pulses Return to Equilibrium

KEY TERMS
Free induction decay Saturated
Longitudinal relaxation Transverse relaxation

Chapter 3 concluded with an explanation of net When a sample is placed in an external mag-
magnetization, vector diagrams, and reference netic field, as indicated by B0 in Figure 4-1, the
frames. Some of this discussion is repeated to nuclei attempt to align with this field. The nuclei
properly develop the origin of the radio signal act like tiny bar magnets, each seeking to orient
that results in the magnetic resonance image. itself with the external magnetic field.
Furthermore, the discussion involves the rotating
frame of reference, not the stationary frame. In MRI, the frequency of precession depends
It is first necessary to consider more closely on the type of nucleus and the intensity of the
the way the magnetic resonance imaging (MRI) external magnetic field.
signal arises. The two basic properties of a
nucleus that are important to this signal are the The spin of the nucleus adds a complicating
nuclear magnetic moment and spin. factor. Instead of each nucleus acting like a simple

49
50 PART I  Fundamentals

results
in M
B0
P B0
µ
Sum of individual spins Net magnetization
vector
FIGURE 4-2  The protons aligned with the external
magnetic field precess randomly at the same frequency.
FIGURE 4-1  In the presence of an external magnetic This results in the net magnetization represented by the
field, protons align with the field and precess at the Larmor vector M.
frequency.

bar magnet, it behaves like a spinning bar magnet. This state of the net magnetization is called
Rather than just aligning with the external field, equilibrium. The proton spins are said to be at
the nucleus precesses around the axis of the equilibrium with the external magnetic field.
external magnetic field, as described by the The magnitude of the net magnetization at
Larmor equation. equilibrium along the Z-axis, symbolized as
M0, is determined by several physical factors.
The more protons available for alignment, the
larger M0 will be.
NET MAGNETIZATION The number of nuclei available that do align
A patient placed in an MRI system consists of a is determined by the intensity of the external
multitude of proton spins. Many of these protons magnetic field. Finally, a large gyromagnetic ratio
attempt to align with the external field and results in a large M0. Therefore N (the number
precess at the Larmor frequency (Figure 4-2). of nuclei available), B0, and γ contribute to M0.
Although all the protons shown are oriented in
an upward direction, the exact direction to which The larger the M0, the stronger the MR signal will
they point at any instant is slightly different be.
because they are at random positions in their
precession. The net result is that the individual Unfortunately, the component of M along the
magnetizations sum to a net magnetization (M) Z-axis cannot be measured directly. It is much
parallel to the direction of the external magnetic too weak compared with B0 to be detected. Only
field. components of the net magnetization vector in
the XY plane, that is, MXY, can be detected by
the MRI receiving coil.
Equilibrium There is no magnetization in the XY plane,
In the preceding situation, the net magnetization MXY. The nuclear spins are all randomly oriented
does not precess but is a vector of constant mag- in these directions; therefore it makes no sense
nitude pointed in the direction of the external to refer to stationary magnetization along either
magnetic field, which is the Z direction. Because the X- or Y-axis. At equilibrium, no signal is
they are randomly oriented, the horizontal or X received from the patient because the net magne-
and Y components of all the individual nuclear tization vector, M, points only in the Z direction,
spins are out of phase, and therefore there is no MZ, and has no component in the XY plane, MXY
net magnetization in the XY plane. (Figure 4-3).
 CHAPTER 4  Equilibrium–Saturation 51

For a signal to be received from a patient, the absorb energy from the RF, and the magnetiza-
magnetization vector must be rotated from the tion vector flips away from the Z-axis
Z-axis so that it has some nonzero component in (Figure 4-4).
the XY plane. This rotation follows a pulse of
radiofrequency (RF) tuned to the nuclei’s Larmor
Flip Angle
frequency. If the RF is not at this frequency, the
nuclei do not absorb energy, and the net magne- The net magnetization vector has been rotated
tization is not rotated. an angle α from the Z-axis because individual
For typical magnetic fields and nuclei of nuclei have absorbed RF energy and are now in
interest, such as hydrogen, the Larmor frequency a high-energy state. The high-energy state is
corresponds to electromagnetic radiation in opposite the direction of the external magnetic
the RF range. Thus if the MRI technologist field. As long as the RF is energized, the net
sends a pulse of RF tuned to the precessional magnetization vector continues to rotate.
frequency of hydrogen, some hydrogen nuclei
The net magnetization vector can be rotated to
any angle by application of a designed time/
intensity RF pulse.
Z
M0

MXY = 0 RADIOFREQUENCY PULSES

MZ = M0
Y Hard and Soft Pulses
B0
The angle through which the net magnetization
X vector is rotated is controlled by two factors.
The speed of rotation is controlled by the strength
FIGURE 4-3  At equilibrium, there is no XY component of the RF pulse. A strong RF pulse rotates net
to net magnetization, and the Z component is at its magnetization rapidly, whereas a weak RF pulse
maximum value, M0. rotates it more slowly.

RF

B0

Z Z α

Y Y

X X
FIGURE 4-4  A radiofrequency (RF) pulse at the Larmor frequency causes the net magnetization vector to flip through
the angle α.
52 PART I  Fundamentals

90° Mz

M
RF M RF RF

MY
M

180°
90° RF pulse 180° RF pulse Partial flip pulse
FIGURE 4-5  In MRI, 90° and 180° radiofrequency pulses are often used to saturate and invert the magnetization,
respectively. Smaller flip angles may be used for fast imaging creating a signal-producing component (MY) with a left-over
longitudinal component (MZ ).

The final angle of rotation is controlled by Net magnetization can be rotated through
the duration of the RF pulse. It is the product any angle. A 90° RF pulse and 180° RF pulse are
of these two factors—the RF pulse intensity most often used in MRI (Figure 4-5).
and duration—that determines the final angle of A 90° RF pulse rotates the net magnetization
rotation. vector from equilibrium onto the XY plane. Sim-
ilarly, a 180° RF pulse rotates the net magnetiza-
Strong, very short RF pulses are called hard tion vector from equilibrium to the-Z-axis. Many
pulses; weaker but longer RF pulses are called fast MRI techniques use so-called “partial flip”
soft pulses. or alpha (α) pulses. The alpha pulse is identified
by a rotation angle less than 90° (e.g., a 10° RF
Thus, for example, a 90° RF pulse can be pulse, a 30° RF pulse).
achieved by either a strong RF pulse of short The use of RF pulses in MRI produces an XY
duration or a weak RF pulse lasting a longer component to the net magnetization. When the
time. Regardless, the duration of even the longest net magnetization vector is rotated, MZ shrinks
RF pulse used in practice is still very short, rarely and MXY grows. As previously mentioned, the
exceeding 10 ms; this is the reason for calling the net magnetization in the XY plane (MXY) is the
burst of RF radiation a pulse. only magnetization that can be detected as an
MR signal from the patient.
XY Magnetization Spins are saturated after a 90° RF pulse so that
Regardless of whether the RF pulse is hard Mz = 0, Mxy = M0.
or soft, what is important is how far the net
magnetization vector has been rotated—the The net magnetization vector shown in Figure
flip angle. For this reason, both hard and 4-6 has absorbed energy from an RF pulse and
soft pulses are most commonly labeled not by is no longer at equilibrium. The components of
their strength and duration but by the angle the net magnetization vector are different from
through which they rotate the net magnetization those at equilibrium. MZ is smaller than M0 and
vector. MXY is no longer zero. When this occurs, the
 CHAPTER 4  Equilibrium–Saturation 53

Excited by RF pulse Larmor frequency of the spins at their local posi-

Z tion in space and the frequency of the rotating
M0 frame (Figure 4-7).
MZ

Y Dephasing of XY Magnetization
MXY Nuclear spins can be influenced by the magnetic
X fields of other nearby atoms. If these fields add
FIGURE 4-6  Flipping net magnetization from the Z-axis to or subtract from the external B0 magnetic field,
reduces MZ and increases MXY. then the Larmor frequency of the spins will
change. If the nearby atoms are moving past the
nuclear spins, like a comet having a close encoun-
ter with a planet, the Larmor frequency will only
nuclear spins are said to be partially saturated. change for a tiny amount of time, but after the
If the net magnetization were totally in the XY magnetic atoms pass by, the nuclear spins will be
plane, then the spins would be saturated. left with a phase shift.
A multitude of these types of interactions
causes the net transverse magnetization to gradu-
MXY Precession ally dephase until there is no signal left at all (see
After excitation by an RF pulse, if the net mag- Figure 4-7). We will discuss this process in more
netization is purely “on resonance,” then in the detail when we talk about T2 relaxation times in
rotating frame the net magnitization vector Chapter 6.
remains staionary; it rotates at the same fre-
quency as the rotating frame itself. However,
Return to Equilibrium
there is no such thing as a perfectly uniform
applied magnetic field. All magnetic fields vary a Immediately after RF excitation, the net magne-
little bit over the volume of the patient, which tization vector seeks to realign itself with the
means that the net magnetization from different external magnetic field. That is, the magnetiza-
parts of the patient has a range of resonant tion vector MZ slowly returns to its equilibrium
frequencies. position as the saturated nuclear spins individu-
Now, if the the net magnetization is in one ally give their energy back to surrounding nuclei
of these “off resonance” regions, then in the in their local environment and return to their
rotating frame the net magnetization vector normal state of alignment with the external mag-
does not remain stationary but rotates. If the netic field. This is shown in Figure 4-8 as the
spins are at a higher frequency they rotate regrowth of MZ along the Z-axis.
counter-clockwise, and if they are at a lower The net result of these two motions, preces-
frequency they will rotate clockwise. sion with dephasing and the return to equilib-
The magnetization vector precesses, just as if rium with B0, are best thought of as two separate
it were an individual nucleus. This occurs because processes (Figure 4-9). The Z component gradu-
the individual nuclear magnetic moments that ally grows until it reaches its maximum value at
were precessing randomly were caused to flip in equilibrium, M0. However, in most tissues the
phase by the action of the RF pulse (see Figure XY component precesses and dephases until the
4-4). The frequency of the rotating frame is net signal disappears at a rate that is at least ten
usually defined by the carrier frequency of the RF times faster than the return of the Z component
pulse for practical purposes. Thus if MXY were to equilibrium.
looked at from above, it would be seen rotating Both of these changes in component magni-
at a frequency that is the difference between the tude, the shrinking of the XY component and
54 PART I  Fundamentals

MXY

  

Spins in phase Precessing, off-resonance Out of phase

MXY

Coherent signal Less signal No signal

FIGURE 4-7  After a radiofrequency pulse that flips net magnetization from the Z-axis, the XY component, MXY, precesses
at a frequency that is the difference between its frequency, ω’, and the Larmor frequency, ωo. In the top row, the solid
line represents on-resonance magnetization that does not move in the rotating frame. The dashed line represents off-
resonance magnetization that precesses at an angular frequency of Δω. If there are groups of spins at a number of
different frequencies they will all eventually dephase (bottom row).

Z M0 component dephases at a much faster rate than

the Z component returns to equilibrium. The
dephasing of MXY is the transverse relaxation
process (or T2); the increasing of MZ is longitu-
dinal relaxation (or T1).

Y In tissue, transverse relaxation occurs more

quickly than longitudinal relaxation.

X
Only MXY, however, is responsible for the MR
FIGURE 4-8  After a radiofrequency pulse, the Z compo- signal. A receiving coil outside the patient views
nent of net magnetization, MZ, relaxes to equilibrium, M0,
along the Z-axis.
the precessional motion as an oscillating mag-
netic field alternately approaching and receding.
The coil is unaware of the return of net magne-
growth of the Z component, are independent tization along the Z-axis to M0.
processes that are linked through the Bloch The electric current induced in the coil by the
equations. The net magnetization vector is not oscillating magnetic field has a waveform like
rigid like a pencil. Rather, it changes size as it that shown in Figure 4-10. This is the primary
rotates, resulting in a complex motion. The XY MR signal: the free induction decay. Secondary
 CHAPTER 4  Equilibrium–Saturation 55

MZ
MZ

No signal, No signal,
magnetization dephased MXY shrinks as MZ grows
FIGURE 4-9  Relaxation of MZ occurs so slowly that the transverse magnetization is totally dephased and the signal is
gone before noticeable recovery occurs. The rate of regrowth of MZ is controlled by the relaxation time, T1.

MXY MXY MXY

Time Time Time

Single frequency, Single frequency, Multiple frequencies,
on-resonance off-resonance off-resonance
FID signal FID signal FID signal
FIGURE 4-10  The radiofrequency (RF) free induction decay (FID) signal due to magnetization at a single resonant fre-
quency looks like a simple decay. A single off-resonance frequency produces a decaying, oscillating signal and when
multiple frequencies contribute to signal interference produces a “beating” pattern.

MR signals, the spin echo and the gradient echo, value, the MR signal is reduced to zero. This
are discussed later. decreasing MR signal, which is received after an
RF pulse, is called a free induction decay (FID).
The free induction decay is the primary MR
FREE INDUCTION DECAY signal.
With all this complex motion of the net magne- When the net magnetization vector is at
tization occurring, what can be observed? The equilibrium, there is no signal. The effect of
only part of the net magnetization that can be the 90° RF pulse is to rotate the net magnetiza-
observed is the XY component. As long as MXY tion vector onto the XY plane. At this point the
precesses with spins in phase and is not zero, an signal decays exponentially. An oscillating MR
oscillating signal is received. The strength of the signal is received from off-resonance spins,
signal received is proportional to the size of the but the different frequencies in the XY compo-
XY component. As MXY dephases to a zero net nent to the net magnetization vector begin to
56 PART I  Fundamentals

90 180 RF pulses transmitted into the patient are

usually indicated on the RFt line by several ovals,
RF pulse RFt as shown on the top line. The label above the
smaller pulse indicates that it is a 90° RF pulse,
and the label above the larger pulse indicates that
RF signal RFs
it is a 180° RF pulse. After a 180° RF pulse, there
is no FID because there is no XY component to
the net magnetization.
FIGURE 4-11  Simple radiofrequency (RF) pulse diagrams
for a 90° and a 180° RF pulse. The top line represents the The signal from the patient is indicated by
transmitted RF pulse (RFt), and the bottom line represents a plot of the intensity of the signal versus time.
the RF signal received (RFs). For the simple case given here, the diagram is
correspondingly simple. In more complicated
situations involving many pulses and signals,
such diagrams are complicated but can be
interfere with each other until there is no signal extremely descriptive.
(see Figure 4-10). Additional lines are added to indicate excita-
tion of the gradient magnetic fields necessary for
The oscillation of the signal is at the same spatial localization of the RFs. For now, it is suf-
frequency as the rotation of Mxy, namely, the ficient to become familiar with this simple two-
Larmor frequency. line form of an RF pulse diagram. This is the
forerunner of the musical score to be discussed
As the net magnetization returns to equilib- in Chapter 16.
rium, the already dephased XY component What about the amplitude and shape of the
shrinks concurrently, and the Z component FID? The amplitude of the FID is equal to the
increases. When MXY relaxes to zero, the MR amplitude of MZ at the start of the RF pulse
signal is again zero. This sequence of events is sequence. This amplitude is often equal to and
shown in Figure 4-9 and, when repeated many always dependent on M0, the equilibrium value.
times, constitutes the simplest MRI RF pulse Therefore the amplitude of the FID is determined
sequence, which is called saturation recovery. by the same parameters that influence M0,
namely, the number of spins involved (the proton
density), B0, and γ.
RADIOFREQUENCY PULSE
DIAGRAMS
CHALLENGE QUESTIONS
In the macroscopic world, two events occur
during an MRI sequence. First, an RF pulse, for 1. Why do different nuclear species (e.g.,
example, a 90° RF pulse, is transmitted into the hydrogen, sodium, fluorine) precess at a dif-
patient. The transmitted RF pulse is symbolized ferent frequency in the presence of the same
as RFt. Second, an RF signal symbolized as RFs magnetic field?
is received from the patient. This signal is the 2. How does one indicate and symbolize
FID, and these two events are diagrammed in an RF pulse that is transmitted into the
Figure 4-11. patient?
There are two lines of information on the 3. At equilibrium, the net magnetization
diagram. The horizontal axis in both cases is vector aligns with the external magnetic field
time. The top line, or RFt, is the RF signal trans- rather than precesses as the proton spins do.
mitted into the patient. The bottom line, or RFs, Why?
is the FID, the MR signal received from the 4. What is the primary MR signal and its
patient. origin?
 CHAPTER 4  Equilibrium–Saturation 57

5. Why can the net magnetization at equilib- 8. What MRI parameter ultimately determines
rium, M0, not be detected and generate an the intensity of the MR signal?
MR signal? 9. What is MZ and MXY for an ensemble of
6. Which takes longer to relax to equilibrium proton spins that are saturated?
value after excitation, longitudinal magneti- 10.  What is the flip angle in MRI?
zation or transverse magnetization?
7. What is the value of MZ and MXY at
equilibrium?
 PA RT
II
The Magnetic
Resonance Image
 CHAPTER

5 
Radiofrequency
Pulse Sequences

OBJECTIVES
At the completion of this chapter, the student • Draw and describe the multi-echo spin echo
should be able to do the following: pulse sequence.
• Identify the components of a radiofrequency • Draw and describe the inversion recovery
(RF) pulse sequence. pulse sequence.
• Draw and describe the partial flip angle pulse • Draw and describe the stimulated echo pulse
sequence. sequence.
• Draw and describe the spin echo pulse
sequence.

OUTLINE
The Basic One-Pulse Sequence Inversion Recovery
Spin Echo and Multi-Echo Stimulated Echo
Spin Echo

KEY TERMS
Inversion recovery Spin echo (SE)
Partial saturation Stimulated echo

Although the free induction decay (FID) is the In most systems the RF transmitter and the
primary magnetic resonance imaging (MRI) signal receiver share the same electronics
signal, it is difficult to use in practice. A hard and frequently the same coil. This makes it
radiofrequency (RF) pulse must be transmitted necessary to switch quickly from transmit
into the patient at the Larmor frequency to mode for the RF pulse to receive mode for
obtain the FID. The FID, which is a weak signal, the FID.
also at the Larmor frequency, immediately Although this switching is done electronically
follows this RF pulse. and is extremely fast, some finite delay is involved.
The RF coil must listen for a very weak This results in the loss of some of the initial part
signal after a very strong excitation RF pulse. of the FID. Therefore, for the patient to provide

59
60 PART II  The Magnetic Resonance Image

TR
90 90 90
RFt

FID FID FID

RFs

FIGURE 5-1  The saturation recovery pulse sequence.

additional signals, more than one RF pulse may sequence to the start of the next pulse sequence
be required. is the repetition time (TR).
This grouping of two or more RF pulses is If the TR is long, on the order of several
called a pulse sequence. The term pulse sequence seconds, the amplitude of the second and succes-
as used in MRI includes the RF pulses and excita- sive FIDs equals the amplitude of the first FID.
tion of gradient magnetic fields. If the TR is short, MZ will not have relaxed to
RF pulses are applied in three basic types of equilibrium, M0. The spins remain partially satu-
pulse sequences in MRI, the one-pulse, two-pulse rated, resulting in a smaller-intensity FID for the
and multi-pulse sequences, which use three or second and successive FIDs. This pulse sequence
more RF pulses. This chapter introduces the is a partial saturation pulse sequence.
reader to the nomenclature that accompanies In order to image more quickly, the flip angle
these pulse sequences. Later chapters detail what can be reduced to α <90°. Using this α-pulse,
these pulse sequences do and how they affect the only part of the longitudinal magnetization, MZ,
magnetic resonance (MR) image. is flipped into the XY plane to produce signal.
This means that there is still some MZ magnetiza-
tion remaining to be used shortly after the first
THE BASIC ONE-PULSE SEQUENCE signal is acquired.
Applying a single RF pulse and producing an In this manner good images can be made using
image from the resulting FID signal is the sim- even short TR values, which allows an increase
plest of pulse sequences. The saturation recovery in imaging speed. The FID signal produced using
pulse sequence was the first of these one-pulse the lower α flip angle RF pulses creates a smaller
sequences. It uses 90° RF pulses and is sometimes signal. However, even with α as low as 10°, a
indicated as 90°-90°-90° … (Figure 5-1). An FID sufficient signal is produced to provide quality
is produced after each 90° RF pulse. images. Therefore in order to use this partial flip
angle pulsing strategy the equilibrium MZ mag-
When Mz = 0 and Mxy = M0, the spins are said netization must be large at the beginning of the
to be saturated. sequence.
This partial flip pulse sequence is indicated as
In MRI, except for some fast imaging tech- α°-α°-α° … . The partial flip pulse sequence is
niques, the data from one pulse sequence are not shown in Figure 5-2 without the gradient mag-
sufficient to generate an image—only a single netic fields.
line’s worth of data. Rather, the sequence must
be repeated many times. It is common for a
SPIN ECHO AND MULTI-ECHO
sequence to be repeated 256 times to obtain data
SPIN ECHO
for a 256 × 256 image.
These repetitions are usually spaced by a MRI uses the FID for very fast techniques;
sizable delay. The time from the start of one pulse however, these do not always produce the most
 CHAPTER 5  Radiofrequency Pulse Sequences 61

Excitation
o o o o

TR

FIGURE 5-2  A pulse sequence that uses a series of α-pulses to produce FID signals.

TR
TE#2
TE#1

90 180 180 90

RFt

FID SE#1 SE#2

RFs

FIGURE 5-3  When the 90° radiofrequency (RF) pulse is followed by an 180° RF pulse, a spin echo (SE) results. This is
an SE pulse sequence.

useful images. If a 180° RF pulse follows the 90° once again is measured from the initial 90° pulse
RF pulse at some later time, an echo signal of the (Figure 5-3). This pulse sequence is a multi-
FID can be generated. echo spin echo (MESE) pulse sequence. In the
This echo signal, called a spin echo (SE), does three-pulse version of MESE each spin echo,
not follow the 180° RF pulse immediately. The acquired at a unique echo time, TE, is used to
spin echo sequence is the most commonly used create an image with special image contrast
two-pulse MRI sequence (see Chapter 6 for characteristics.
further discussion). Now if we continue to apply 180° pulses with
the same time delay that was used between the
The time between the initial 90° RF pulse and first two 180° pulses, we will generate a train of
the SE is called time-to-echo (TE). spin echo signals at later and later values of TE.
We could use each of these later spin echo signals
The time between the 180° RF pulse and the to generate an entirely new image; however, radi-
SE equals the time between the 90° and the 180° ologists have found these very long TE images to
RF pulses. By controlling the time for the 180° be of little diagnostic value.
RF pulse, the MRI technologist can control TE. A different strategy is to use each of these later
The SE pulse sequence is indicated as 90°-180° echo signals to produce a different line in the
… 90°-180° … . image. This speeds up the spin echo imaging
If another 180° RF pulse follows the SE, an process and is the basis for fast spin echo imaging,
additional SE can be generated with a TE that which will be discussed in more detail in Chapter
62 PART II  The Magnetic Resonance Image

TR
TE#2
TE#1
TI
180 90 180 180 180

RFt

FID SE#1 SE#2

RFs

FIGURE 5-4  The inversion recovery pulse sequence also produces spin echoes (SEs).

16. The three-pulse MESE pulse sequence is indi-

STIMULATED ECHO
cated as 90°-180°-180° … 90°-180°-180° … .
When three RF pulses are used and if the time
between the pulses is less than T2, then a differ-
INVERSION RECOVERY ent kind of echo signal, called the stimulated
Another common pulse sequence using three echo (STE) can be produced. This is most easily
RF pulses is the inversion recovery (IR). The IR understood if we consider what happens if three
RF pulse sequence uses a 180° RF pulse before 90° RF pulses are applied in a rapid series (Figure
the 90° RF pulse. This 180° RF pulse rotates 5-5). The first RF pulse produces an FID signal,
the net magnetization vector through 180° as we know from our previous discussion. During
onto the –Z-axis; that is, it inverts the net the time between pulses the transverse magneti-
magnetization. zation dephases and then a second 90° RF pulse
is applied. The component of the transverse mag-
No MR signal is generated after a 180° RF pulse. netization that is parallel to the direction of the
RF pulse produces a spin echo. However, the
component of the transverse magnetization that
After a delay called inversion-delay-time (TI), is perpendicular to the direction of the RF pulse
a 90° RF pulse is applied to rotate the net mag- is rotated to become longitudinal magnetization.
netization onto the XY plane where it can be During the next interpulse period the spin echo
detected and produce a signal. Once again, SEs refocuses, but the MZ magnetization does not
are created by adding one or more 180° RF dephase. When the third 90° RF pulse is applied
pulses (Figure 5-4). the MZ magnetization is rotated back into the
As with the SE pulse sequence, the FID transverse plane and an STE signal is produced
produced by the 90° RF pulse is disregarded. (Figure 5-6).
Only SEs are used to make an image. The IR The STE is only ocassionally used to produce
pulse sequence is indicated as 180°-90°-180° … MR images. If the pulses are not all perfect 90°
180°-90°-180° … 180°-90°-180° … . Subsequent pulses then the rapid application of three RF
chapters will expand on the importance of TR, pulses actually creates not one spin echo but four,
TE, and TI in image production and contrast so that the total number of echo signals being
rendition. generated is five. Therefore it is easy to generate
 CHAPTER 5  Radiofrequency Pulse Sequences 63

X Y X Y X Y X Y

90° pulse 90° pulse

t1

Spin
echo

X Y X Y X Y X Y
Stimulated
echo
90° pulse

t2 t1
FIGURE 5-5  A stimulated echo (STE) in the rotating frame occurs after the 1st RF pulse rotates MZ into the XY plane
and the magnetization partially dephases. The 2nd RF pulse refocuses some of MXY as an SE but also rotates some of it
back to MZ. The third RF pulse rotates the MZ back to the XY plane where it refocuses to form a stimulated echo (STE).

90° 90° 90°

t1 t2 t1

SE STE
FIGURE 5-6  A pulse sequence will produce stimulated echoes (STE) if three pulses are applied in rapid succession. In
the simple case where all three pulses have a 90° flip angle, an SE and an STE will be produced.
64 PART II  The Magnetic Resonance Image

unwanted stimulated echoes in very fast pulse

CHALLENGE QUESTIONS
sequences, and they can be a major source of
artifacts. This is why it is important to know 1. What makes the FID more difficult to detect
about the STE signals so that the problems they than an SE?
cause can be recognized. 2. Describe the RF pulse sequence used for SE
The following summarizes these principal imaging.
MRI RF pulse sequences: 3. What does the term pulse sequence mean
when applied to MRI?
MRI RF Pulse Sequences 4. Of all the available MRI pulse sequences,
which provides the fastest imaging?
SR 90° . . . . . . 90° . . . . . . 90° . . . . . . 5. How is the RF pulse indicated for an STE
Partial Flip a ° . . . . . . a ° . . . . . . a ° . . . . . . echo pulse sequence?
6. What is the TE in SE imaging?
SE 90° - 180° . . . . . . - 90° - 180° . . . . . . 7. Diagram the transmitted and received signals
90° - 180° . . . . . . for a double echo, SE pulse sequence.
8. What is the inversion time in an IR pulse
IR 180° - 90° - 180° . . . . . . 180° - 90° - 180° . . . . . .
sequence?
STE a ° - a ° - a ° . . . 9. What is the MRI signal called when it is
obtained during an IR pulse sequence?
10. What is the minimum number of pulses
required to produce an STE?
 CHAPTER

6 
Magnetic Resonance
Imaging Tissue Parameters

OBJECTIVES
At the completion of this chapter, the student • Discuss phase and phase coherence.
will be able to do the following: • Explain the difference between T2 and T2*
• Identify the three principal magnetic resonance relaxation times.
imaging (MRI) tissue parameters.
• Recite the various ways for identifying T1 and
T2 relaxation times.

OUTLINE
Proton Density T2 Relaxation Time
T1 Relaxation Time Phase Coherence
T1ρ Relaxation Time T2* Relaxation

KEY TERMS
Dephase Spin ensemble
Phase coherence Susceptibility
Proton density

It is essential to understand the following three Images of each of these parameters can only
principal parameters that are inherent properties be produced indirectly, using image processing
of biological tissues and which make possible the methods to calculate these parameters from the
soft tissue image contrast, for which MRI is data from several acquired images. However,
known. These three inherent parameters of mag- images weighted by each of these parameters are
netic resonance imaging (MRI) are proton density routinely produced by carefully selecting the type
(PD), T1 relaxation time, and T2 relaxation of radiofrequency (RF) pulse sequence and the
time. Each is fundamentally different from and RF pulse times, which include repetition time
independent of the others. (TR), time-to-echo (TE), and inversion-delay

65
66 PART II  The Magnetic Resonance Image

time (TI). Such images are identified as proton Z

density weighted (PDW), T1 weighted (T1W),
and T2 weighted (T2W) according to which
Y
parameter is emphasized.
X bone

PROTON DENSITY
The amplitude of the net magnetization is related Z
to several parameters, as previously described in Z
Chapter 3. It is reasonable to expect that one of
Y
the parameters that affects the amplitude of the
Y blood
magnetic resonance (MR) signal might be the fat X
number of hydrogen nuclei within the volume of X
the sample. If no hydrogen nuclei are present, no FIGURE 6-1  M0 and signal intensity are proportional to
signal should be expected. Conversely, if the mobile hydrogen concentration or proton density (PD).
These three voxels each have different PDs.
sample is rich in hydrogen, a strong MR signal
may be expected. Voxels with high hydrogen
concentration or proton density (PD) appear
bright. This reasoning holds true only to a certain PD is the measure of the concentration of
extent. mobile hydrogen nuclei available to produce an
MR signal. The higher the concentration of
MR signal intensity is proportional to PD, γ2, mobile hydrogen nuclei, the stronger the net
and B0. magnetization at equilibrium (M0) and the more
intense the MR signal. A strong MR signal results
MRI signal amplitude depends on the pres- in a better MR image.
ence or absence of hydrogen nuclei and is also Figure 6-1 shows three pixels highlighted in
sensitive to the environment of the hydrogen the cross section of the patient. Each represents
nuclei. How hydrogen is bound within a mole- a voxel containing different tissue with different
cule also influences the amplitude of the MR concentrations of mobile hydrogen nuclei. The
signal. tissue with the highest proton density (PD) also
The signal from tightly bound hydrogen has the largest value of M0. Table 6-1 shows
dephases and disappears very quickly. Therefore averages of relative values of PD for several
the typical MR signal received originates from tissues.
loosely bound hydrogen nuclei, which are called Although a large M0 is necessary for a strong
mobile hydrogen. Hydrogen nuclei found in MR signal, the MZ is undetectable. In an MRI
liquids are mobile. system, the receiver detects only the XY compo-
An example of this effect is bone. Cortical nent of net magnetization—MXY. Because MXY is
bone appears black on an MR image because zero at equilibrium, no MR signal is received.
its MR signal is totally dephased before the Patients whose nuclei are at equilibrium do not
receiver is turned on. This is not due to the emit an MR signal. The net magnetization must
absence of hydrogen; rather, the hydrogen nuclei have a component in the XY plane to produce
are tightly bound to the molecule. As a result, an MR signal.
bone typically looks like air, which has a low
hydrogen concentration. However, medullary
T1 RELAXATION TIME
bone is visible because of the fat located in the
spaces between the trabeculae and in the marrow When a patient is positioned in the magnetic field
cavities. of an MRI system, the net magnetization of the
 CHAPTER 6  Magnetic Resonance Imaging Tissue Parameters 67

patient does not change instantly. In fact, the MZ continues indefinitely unless the spins are dis-
changes rapidly at first and then relatively slowly turbed. If the magnet is turned off or the patient
(Figure 6-2; see also Appendix A). The longitu- is removed from the magnetic field, the individ-
dinal magnetization, MZ, reaches equilibrium ual nuclear magnetic moments reposition ran-
after approximately five T1 relaxation times. domly. This causes the net magnetization along
the Z-axis to disappear, so MZ returns to zero
One T1 relaxation time results in relaxation to (Figure 6-3).
63% of M0. The time constant that describes the rate at
which MZ returns to M0 is the T1 relaxation
Once equilibrium is reached and MZ time. For tap water, T1 is about 2500 ms. For
equals M0, this condition of net magnetization tissue in vivo, T1 can be as short as 100 ms for
fatty tissues and as long as 2000 ms for bodily
fluids such as cerebrospinal fluid (Table 6-2).
TA B L E 6 - 1 Approximate Values of
Mobile Hydrogen Nuclei
Spin Density for Various Generally, the T1 of diseased and damaged tissue
Tissues is longer than that for corresponding healthy
tissue.
Tissue Relative Spin Density
An equation describing how MZ relaxes
Muscle 90
White matter 60 toward equilibrium has an exponential form.
Fat 95 The further MZ is from M0, the faster it
Cerebrospinal fluid 100 approaches M0. As the ensemble of nuclear spins
Kidney 95 gets closer to M0, MZ approaches more slowly.
Gray matter 70 The ensemble of nuclear spins relaxes to
Spleen 90 0.63 M0 in one T1. As a result, it takes approxi-
Liver 90 mately five T1 for a spin ensemble to return to
Blood 85 M0 once it has been disturbed. When repetitive
Cortical bone 1-10
MR signals are sampled for imaging, the amount
Lung 1-5
Air <1 of magnetization available for the next sampling
is limited.

M0

Equilibrium

0.63M0

MZ

0
1T1 2T1 3T1
FIGURE 6-2  The relaxation of MZ with time after positioning a patient in a strong magnetic field depends on the T1
relaxation time.
68 PART II  The Magnetic Resonance Image

maximum amplitude, M0. If the nuclei become

M0
disturbed by directing an RF pulse at the Larmor
frequency into the patient, the nuclei absorb
MZ energy, and the net magnetization is rotated and
changed in magnitude.
0.37M0
This new magnetization state of the patient is
unstable because the nuclei want to realign with
0
1T1 2T1 3T1 the external magnetic field and return to equilib-
Time rium. The regrowth of MZ is the T1 process and
FIGURE 6-3  MZ shrinks with time constant T1 when the represents the spin’s return to equilibrium.
patient is removed from an external magnetic field. Nuclei must give up the energy gained from
the transmitted RF pulse to return to equilib-
rium. If the hydrogen nuclei are bound with
TA B L E 6 - 2 Approximate T1 Relaxation other atoms to form a molecule, this arrange-
Times (in ms) at Three B0 ment is called a lattice. In biological tissues, only
Field Strengths for Various bones have atoms bound in lattice structures, but
Tissues this terminology is typically applied to all nuclear
spins that interact with the magnetic fields of
0.35 1.5 3.0 other atoms in the cell.
Tissue Tesla Tesla Tesla The regrowth of MZ is accomplished by trans-
Adipose 200 260 290 ferring energy to other atoms in the molecule or
Liver 285 495 640 to the molecule as a whole; that is, the energy
Renal cortex — 1000 1150 is released by an interaction between the indi-
Renal medulla — 1300 1550 vidual hydrogen nuclei (spin) and the surround-
Spleen 485 780 985 ing molecules (lattice). Pixel intensity in an MR
White matter 500 650 800
image is a complicated function of the T1 relax-
Gray matter 800 1200 1500
Muscle 500 1000 1400 ation time. Whether a pixel appears bright or
Heart muscle 525 870 1100 dark depends on the RF pulse sequence.
Oxygenated blood* 900 1350 1650
Deoxygenated blood* 900 1350 1575 Generally, for T1W images, tissue with short T1
Cerebrospinal fluid 3300 3350 3400 appears bright; tissue with long T1 appears dark.
Water 3500 3500 3500

*At typical human hematocrit of 0.42.

T1ρ Relaxation Time
T1 is also called the longitudinal or spin- The magnitude of MZ and MXY after the applica-
lattice relaxation time. The term longitudinal tion of a 90° RF pulse is shown in Figure 6-4.
refers to events occurring along the axis of the They are similar in shape. Much like radioactive
net magnetization vector, which is parallel to B0. decay, they are both exponential in form. The T1
T1 relaxation describes the growth and decay of relaxation time is much longer than the T1ρ
MZ; therefore it is longitudinal in nature. relaxation time. Just as T1 relaxation controls
MZ, T1ρ relaxation controls MXY. The constants
  T1 relaxation time is the same as longitudinal T1 and T1ρ control the rate of relaxation of MZ
relaxation time or spin-lattice relaxation time. and MXY, respectively. Each is a fundamental
property of tissue.
When a patient is in the magnet at equilib- T1ρ is less well-known than T1 and T2, but
rium, the net magnetization is constant and of it represents a specific type of relaxation process
 CHAPTER 6  Magnetic Resonance Imaging Tissue Parameters 69

MO very restricted. T2 processes cause a rapid decay

of the MRI signal (as we discuss below) and the
MZ
T1 recovery T1ρ decay of the magnetization, under the strong
influence of low-frequency interactions, proceeds
A more slowly than T2 but much more rapidly
than T1.
MO
Under most conditions in biomedical MRI,
T1 decay T1ρ is not important because by the time it
MXY
would exert a strong influence on the evolution
T2 dephasing of the magnetization, the T2 and T2* processes
have totally obliterated the MRI signal. However,
*
T2 dephasing
B in bones and cartilage using special pulse
FIGURE 6-4  The relationships between the various types sequences, the use of T1ρ contrast is becoming
of relaxation times are shown here. A, T1 characterizes the more common. T1ρ effects can be used to influ-
amount of time for the transverse magnetization to reach ence MRI contrast using a special “spin locking”
equilibrium. B, T1ρ characterizes the reduction in the mag-
pulse sequence (Figure 6-5).
nitude of MXY, which is usually somewhat faster than the
rate of MZ growth. T2 does not affect the magnitude of
individual components of the MXY vector, but produces a
net reduction due to dephasing of MXY (loss of coherence)
T2 RELAXATION TIME
by means of microscopic interactions among various mol- A large value for T1 or T2 indicates a long,
ecules. T2* is an additional dephasing caused by various
gradual relaxation; a small value indicates a
macroscopic B0 inhomogeneities.
rapid relaxation. This occurs much like a struck
bell, which reverberates for several seconds.
It may seem strange that T1 relaxation is
that is becoming more important in clincal represented by an increasing value with time,
imaging. The “rho” in T1ρ stands for the “rotat- whereas T2 relaxation is represented by a
ing frame” because T1ρ is the T1 relaxation in decreasing value. This happens because the relax-
the rotating frame. It is represented by the ation of MZ to M0 is actually a relaxation from
reduced amplitude of the MXY magnetization the excited state to the equilibrium state, whereas
while MZ increases (see Figure 6-4). MXY relaxation describes a loss of phase
Because the relaxation of MZ is independent coherence.
of MXY, T1 and T1ρ are also independent. T1ρ
is always less than T1. Because the MR signal T2 relaxation never exceeds T1 relaxation time
received is proportional to MXY, it also relaxes for any given tissue.
according to T1ρ, and the MXY magnetization
decreases while MZ increases. Analogous to T1, T2 is called the transverse
In liquids such as pure solutions, T1ρ is almost relaxation time. Because T2 relaxation repre-
identical to T1 and T2, because the rate of MZ sents a loss of net XY magnetization, it repre-
increasing due to T1 is almost the same as the sents the loss of phase coherence in a plane
rate that MXY decreases due to T1ρ. For example, perpendicular to or transverse to M0. M0 lies
the T1, T1ρ, and T2 for pure water are all along the Z-axis.
approximately 2500 ms, indicating that the
signal would take several seconds to relax. T2 relaxation time is the same as transverse
That is contrasted by the behavior in the relaxation time or spin-spin relaxation time.
solids, where there is a great difference in the
three relaxation times. In a pure crystalline sub- In any material above absolute zero in tem-
stance, for instance, the motion of the atoms is perature, molecules and atoms are in constant
70 PART II  The Magnetic Resonance Image

MZ MZ
Equilibrium First 90°
magnetization in RF pulse
rotating frame

X X
Y Y

Spin-lock
condition
Effect of
first 90°
RF pulse
B1Y

MXY

B1X X
X MY
Y
Y

FIGURE 6-5  T1ρ measurements are performed by applying two RF pulses in succession. The first is applied along the
Y’-axis (a conventional 90° excitation pulse) and immediately thereafter a second RF pulse is applied on the X’-axis. The
second RF pulse “spin locks” the transverse magnetization to the X’-axis in the rotating frame. During the “spin locking”
the transverse magnetization cannot dephase but the length of the net magnetization decreases as a function of T1ρ.
This provides a unique way to probe the low-frequency molecular interactions.

motion. At room temperature, this motion is so

TABLE 6-3 Approximate T2 Relaxation
rapid that the magnetic fields of the nuclei con-
Times (in ms) at Three
tinually interact with each other. B0 Field Strengths
These interactions cause the magnetic field for Various Tissues
that one hydrogen nucleus in a water molecule
experiences to change for a very short period 0.35 1.5 3.0
of time so that it is slightly greater than or less Tissue Tesla Tesla Tesla
than the magnetic fields at other nuclei. As a
Muscle 45 45 50
result, the resonance frequency changes momen- Liver 65 50 40
tarily and causes a small phase shift. This is Renal cortex 55 70 90
the principal mechanism for the relaxation of Renal medulla 140
MXY—transient changes of the magnetic fields Spleen 80
experienced by one spin being different from Adipose 60
others. White matter 110 75 50
Pixel intensity in an MR image is also a Gray matter 125 100 75
complicated function of T2 relaxation time. The Oxygenated blood* 240 167 120
Deoxygenated blood* 210 55 25
pulse sequence that is used determines the
Cerebrospinal fluid 2200
T2-related brightness of pixels. Table 6-3 pre­ Water 3500
sents approximate values of T2 relaxation times
for various tissues. *At typical human hematocrit of 0.42.
 CHAPTER 6  Magnetic Resonance Imaging Tissue Parameters 71

T2 relaxation

T2* relaxation 1 2 3

FIGURE 6-6  Theoretically, MXY should relax at the same

rate as T2. The actual relaxation is much shorter and is FIGURE 6-7  To visualize T1 and T2*, consider a patient
called T2*. in a magnet whose trunk is divided into three regions.

Generally, with T2W images, tissue with long T2 of the three regions are flipped onto the XY plane
appears bright; tissue with short T2 appears dark. along the Y-axis (Figure 6-8).
Even though the patient is considered to be
divided into three regions, the received MR
Phase Coherence
signals still come from throughout the patient.
With the three fundamental MRI parameters Therefore the signal is the vector sum of the net
known, a simple experiment can be done to see magnetization of the three regions. Because the
whether the effects of each can be observed. After magnetization vectors in the three regions all
a 90° RF pulse, the net magnetization is rotated point in the same direction, they initially differ
onto the XY plane. A large MXY will be detected only due to the proton density of each tissue and
as a signal. As a result of T1 relaxation, because add maximally and produce a large signal.
MZ grows then MXY must shrink and the signal If only the T1 effects are considered, then at
will shrink, too. Before equilibrium is reached, some later time (time A), the three net magnetiza-
the MXY will relax to zero. tion vectors of Figure 6-8 have changed. They
However, unexpected results are obtained are shorter because of T1 relaxation. If they are
when the experiment is performed. After excita- off-resonance, they have rotated and now point
tion with a 90° RF pulse, the FID shrinks more in a new direction (i.e., they are precessing). The
rapidly instead of lasting for several seconds. The rate of precession is the same for each net mag-
relaxation time obtained from such a measure- netization vector because this is controlled by the
ment is called T2 star (T2*). T2* is much shorter magnetic field strength in the three regions and
than T2 (Figure 6-6). In order to explain our B0 is the same in each region.
results we need to consider the three similar, but The net result is that each MXY is smaller and
different processes that are occurring. has changed direction, but they all still point in
Consider the nuclei in the B0 magnetic field of the same direction. Their sum is smaller than at
a patient who is approximately 50 cm long. We time zero because of T1 relaxation.
will look at the signals from three tissues (i.e., However, the T2 relaxation process occurs
liver, renal cortex, and renal medulla) and con- much more quickly than T1. Before there is a
sider the nuclear spins that lie within each of significant reduction in MXY the phase of the
these three regions (Figure 6-7). After the 90° RF spins that contribute to the total MXY vector
pulse at time zero, the net magnetization vectors begin to dephase due to the multitude of
72 PART II  The Magnetic Resonance Image

In xy plane
IN PHASE
Region Region Region Total
1 2 3 magnetization

Time 0 Y Y Y Y

X X X X

Time A Y Y Y Y

X X X X

Time B Y Y Y Y

X X X X
FIGURE 6-8  In a perfect magnet, loss of MXY is the same throughout a homogeneous tissue.

0 each have the same frequency (i.e., one revolu-

T2 relaxation
tion per hour for the minute hand) and the same
B phase (same time), because all hands point in the
Signal same direction. A reception desk in a fancy San
C Francisco hotel also has five clocks (Figure 6-11).
A
These clocks indicate the time in various cities
around the world. The clocks have the same
FIGURE 6-9  The envelope of the free induction decay frequency, but they are out of phase.
obtained from a perfect magnet describes T2 relaxation.
Times 0 (zero), A, B, and C refer to Figure 6-8. The key assumption in the earlier description
is that the magnetic field in the three regions of
the patient is the same. Such B0 field uniformity
interactions with the magnetic fields of other causes the net magnetization vectors to dephase
molecules. Now they point in various directions at a rate that is determined solely by the micro-
as a result of precession and their vector sum is scopic interactions that go on in each type of
reduced due to the dephasing process. tissue.
The dotted line in Figure 6-9 represents the It is impossible to build such a perfect magnet
envelope of the dephasing, precessing signal, in real life. Although the magnets used in MRI
which results in an FID. The envelope is reduced are high quality, the magnetic fields are not per-
with a time constant T2. This condition in which fectly uniform. The field strength varies slightly
all the dephasing of spins precess differently due from place to place in the imaging aperture. Even
to multiple, microscopic interactions with their this small nonuniformity in magnetic field inten-
local enviroments is termed out of phase. The sity greatly affects the MR signal.
spins have lost their phase coherence.
Frequency and phase are sometimes confused.
T2* Relaxation
For example, a saloon in Houston has five clocks
above the bar (Figure 6-10). Each indicates the A different FID results if the earlier MRI experi-
time in a city in the central United States. They ment is repeated with a less-than-perfect magnet.
 CHAPTER 6  Magnetic Resonance Imaging Tissue Parameters 73

all parallel and add maximally as before.

However, as time progresses, something different
occurs. The magnetization vectors decay much
Chicago Dallas Houston Kansas City Happy Hour more rapidly.
Each vector shrinks by the same amount as a
result of T2 relaxation; however, because the
magnetic field is slightly different in the three
regions, each vector rotates differently so that
they now point in slightly different directions.
For example, the magnetization vector in region
1 is in a slightly higher field and precesses at a
slightly faster rate; therefore, at time A, it has
rotated a little further than its neighbor in region
2. In contrast, the magnetization vector in region
3 is in a slightly lower magnetic field and there-
fore has precessed more slowly. It points in yet
another direction.
When these three magnetization vectors are
FIGURE 6-10  The five clocks above the bar in a Houston added together, the total net magnetization vector
saloon have the same frequency. They are also in phase.
is smaller for two reasons. First, T2 has short-
ened all three net magnetization vectors due to
dephasing induced by microscopic interactions.
Second, the vectors have additional dephasing
due to the macroscopic inhomogeneities in the B0
magnetic field. Therefore the total net magnetiza-
tion vector has dephased more and is shorter
than that for the perfect magnetic field at any
particular point in time.
At time B, the effect of the variations in
the magnetic field on the direction of the magne-
tization vectors is even more pronounced. The
higher-field magnetization vector on the left is
farther ahead, and the lower-field magnetization
vector on the right falls farther behind. The sum
is almost zero.
The resulting MR signal relaxes more rapidly
than that caused by T2 alone (Figure 6-13). The
small variations in the magnetic field cause the
FIGURE 6-11  The five clocks above the reception desk magnetization vectors in different regions to
of a fancy San Francisco hotel have the same frequency, precess at different frequencies; that is, the spins
but they are out of phase. in different regions rapidly lose their phase
coherence—they dephase.
Once again, a 90° RF pulse is transmitted into Although this discussion has used a three-
the patient, and the net magnetization vectors in compartment model, the actual situation involves
the three regions flip together onto the XY plane a multicompartment model. The scale of the
along the Y-axis, as shown at time zero in Figure magnetic field inhomogeneity is not regional but
6-12. At this point, the magnetization vectors are intravoxel.
74 PART II  The Magnetic Resonance Image

In XY plane
DEPHASED
Region Region Region Total
1 2 3 magnetization

Time 0 Y Y Y Y

X X X X

Time A Y Y Y Y

X X X X

Time B Y Y Y Y

X X X X
FIGURE 6-12  In a magnetic resonance imaging system, inhomogeneity of the magnetic field causes MXY to relax more
rapidly than expected. The result is T2* relaxation.

0 differences between some tissues (i.e., their sus-

T2* relaxation
ceptibilities) can also be an important determi-
B nant of T2*.
B*
Signal
To summarize, the relationships between the
A C various types of relaxation times are shown in
A* Figure 6-4. T1 characterizes the amount of time
T2 relaxation for the transverse magnetization to reach equi-
FIGURE 6-13  The envelope of the free induction delay librium. T1ρ characterizes the reduction in the
obtained from an imaging magnet describes T2* relax- magnitude of MXY, which is usually somewhat
ation. T2* is considerably less than T2 because of mag- faster than the rate of MZ growth. T2 does not
netic field inhomogeneities. affect the magnitude of individual components
of the MXY vector, but produces a net reduction
due to dephasing of MXY (loss of coherence)
by means of microscopic interactions among
The inhomogeneity in the magnetic field causes various molecules. T2* is an additional dephas-
each spin to precess differently so that T2* is ing caused by T2 mechanisms plus macroscopic
significantly shorter than T2. B0 inhomogeneities.

The T2* measured in such an MRI experi-

CHALLENGE QUESTIONS
ment combines two factors: the real T2 of
tissue and the magnetic field inhomogeneities. 1. Name the three principal MRI parameters
Typically, in practical magnets, the effect of characteristic of each tissue.
magnetic field inhomogeneities is about the 2. Arrange the three MRI relaxation times, T1,
same as that of T2 relaxation. However, we will T2, and T2*, from shortest to longest for
also see in future chapters that magnetic field soft tissue.
 CHAPTER 6  Magnetic Resonance Imaging Tissue Parameters 75

3. Which MRI characteristic of tissue princi- 6. Distinguish between frequency and phase.
pally determines the intensity of the MR 7. T1 relaxation is spin-lattice relaxation. To
signal? what does lattice refer?
4. What is the principal reason that the T2* is 8. What does the envelope of an FID represent?
always shorter than T2? 9. In a T1W image, which tissues appear bright
5. How much longitudinal relaxation occurs and which dark?
during one T1, and approximately how 10. When a T2W image is presented, what
many relaxation times are needed for com- tissues will appear bright?
plete relaxation to reach equilibrium?
 CHAPTER

7 
Manipulating Magnetic
Resonance Image Contrast

OBJECTIVES
At the completion of this chapter, the student • Plot T2 relaxation and its relation to T2*
should be able to do the following: relaxation.
• Identify the radiofrequency (RF) pulse • Identify the RF pulse sequences required to
sequence used to produce T2-weighted images. produce T1-weighted images.
• Identify the RF pulse sequences required to • Explain how MRI contrast agents are used to
produce proton density–weighted images. change relaxation times.
• Identify the RF pulse sequences required to
produce proton T2*-weighted images.

OUTLINE
How to Make T2-Weighted How to Make Proton Density– T1-Weighted Images versus
Images Weighted Images T2-Weighted Images
How to Make T2*-Weighted How to Make T1-Weighted Contrast Agents in MRI
Images Images

KEY TERMS
Contrast agents Inversion recovery Magnetic field inhomogeneity
Dynamic range Inversion time Rephase

Basic magnetic resonance image contrast is density or proton density, longitudinal relaxation
affected by the amplitude and timing of the RF time (T1), and transverse relaxation time (T2).
pulses used to excite the spin system. More There are several other important intrinsic
advanced methods may use gradient pulses (to tissue parameters that will be considered in later
modulate motion) and alter tissue properties chapters. These include chemical shift, which
with exogenous contrast agents. Previously we is the difference in the resonant frequency
have discussed the main intrinsic, tissue MRI between proton nuclei in water and proton
parameters that affect image contrast: water nuclei in fat molecules. Tissue motion, which

76
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 77

includes both macroscopic motion (breathing,

HOW TO MAKE T2-WEIGHTED
blood flow, peristaltic motion) and microscopic
IMAGES
motion (perfusion and diffusion), is an additional
important consideration. Also we will take into The radiofrequency (RF) pulse sequence shown
account differences in the magnetic susceptibility in Figure 7-1 results in a magnetic resonance
of various tissues. (MR) signal, the spin echo (SE). The FID does
Although the properties of the tissues are not relax according to the T2 of the sample
important, the contrast in MR images is con- because of B0 inhomogeneity; so how can we
trolled by a number of user-selectable parame- make the signal truly dependent on T2? Several
ters, which are predetermined by the technologist methods have been developed, all of which use
at the MRI operating console. Magnetic field additional pulses after the initial 90° RF pulse.
strength is also important, but this is determined The spin echo pulse sequence is the most common
when the equipment is purchased and cannot be of these (see Chapter 4).
changed. In Figure 7-2, the net magnetization in the
Image contrast in MRI is principally con- three regions of the patient is represented by
trolled by the timing of radiofrequency (RF) three runners on an oval track. At time zero, they
pulses using specific delays (TR, TE, TI, etc.) all leave the starting line together. A perfect
and modifying the amplitudes of these pulses magnet system is represented by runners moving
to control the flip angles. A large number of at exactly the same speed, so that at some later
other parameters also influence contrast in time the runners are still exactly together and in
specific circumstances, including the amplitude step (Figure 7-3).
and timing of the gradients, the excitation fre- However, in a real magnet the net magnetiza-
quency of the RF pulses, the bandwidth of the tions in the three regions dephase because spins
RF pulses, and the bandwidth of the MRI receiver are precessing at slightly different rates (Figure
subsystem. 7-4). The runners start together but each runs at
All of these parameters will be considered a slightly different speed, so that after some point
in following chapters. In this chapter, however, in time, they are no longer together.
we will focus on the relaxation times that we Is there some way to cause the runners to
previously discussed and how the timing and come back together even though they each run
amplitudes of the RF pulses can be used to at a different speed? It is possible if a new rule
exploit them to maximize soft tissue image
contrast.
Although the following section describes
methods for modifying image contrast based on 90° (Excitation) 180° (Rephasing)
the relaxation times, we should note that it is
also possible to measure the relaxation times TX Time
directly and produce images of the relaxation
times (called parametric maps). This approach
FID Spin Echo
is used primarily in research since it is time-
RX
consuming and computationally intensive. Time
However, recently some studies using injected
contrast agents have included relaxation time
TE
maps, and this topic will be considered in
Chapter 27. In current clinical practice, relax-
FIGURE 7-1  Spin echo pulse sequence diagram shows
ation times are not measured or calculated; how the FID and spin echo signals decay as exp(-T2*/t),
they are referred to as relative weighting of while the peak signal of the spin echo is reduced from the
an image. peak signal of the FID by an amount exp(-TE/T2).
78 PART II  The Magnetic Resonance Image

is introduced to the race. For example, at the

time in Figure 7-4, a whistle is blown, and all the
runners immediately have to turn around, revers-
ing their direction, and run back toward the
starting line.
If the race is now run making use of this rule,
the scene will change. The runners start out
together but soon begin to separate, that is,
dephase. Now at time A, the whistle is blown,
and all the runners reverse and start running
back toward the starting line. Suddenly the
fastest runner, who was far ahead, is behind, and
the slowest runner, who was behind, is in the
lead. Even though they have changed directions,
they have not changed speed. Therefore even
though the faster runner is now behind, he will
catch up with the others.

A 180° RF pulse causes spins to rephase and

form a spin echo.

FIGURE 7-2  Immediately after a 90° RF pulse, all spins Running toward the starting line is exactly the
begin at the same starting line, just like runners in a race. reverse of the start, when the runners ran from

FIGURE 7-3  A perfect magnet is analogous to a perfect race. In a perfect magnet, MXY remains in phase. In a perfect
race, the runners remain in step.
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 79

FIGURE 7-4  In an imaging magnet, MXY dephases rapidly. In a real race, competitors run at different speeds.

the starting line. This means that as the runners

cross the starting line (Figure 7-5), they will
again be precisely together; they are in phase
again. They will cross the starting line at a time
exactly twice that when the whistle was blown.
For example, if the whistle was blown at 20
seconds, “rephasing” of the runners occurs at 40
seconds.
If the race continues, the runners again dephase
(Figure 7-6). However, they can then be forced
to rephase by blowing the whistle again, which
will cause them to reverse direction again and
come together at the starting line.
This process can be repeated any number of
times. Each time the runners will rephase, and
the information lost because of the difference
in runners’ speeds is recovered. The difference
in the runners’ speeds is analogous to the dephas-
ing that occurs as a result of magnetic field
inhomogeneities.
In a magnet, the runners (also known as spins)
start out going both clockwise and counterclock- FIGURE 7-5  If the runners in Figure 7-4 reverse direction,
wise (Figure 7-7) and the reversals of the runners they will cross the starting line at precisely the same time.
are accomplished with 180° RF pulses. Each They are back in phase.
80 PART II  The Magnetic Resonance Image

FIGURE 7-6  After additional time, the runners are again out of phase but not running quite so fast. A whistle blown
at this time causes the runners to reverse direction again and rephase at the starting line.

180° pulse causes all the net magnetization additional spin echoes. Each additional spin echo
vectors to flip 180°, in essence, reversing their will be reduced in amplitude.
direction. The spins then begin to rephase, and The spin echo first increases in intensity to a
as they do, a signal is generated. The maximum maximum, and then relaxes back to zero (see
signal occurs at the point where they are again Figure 7-1). The first half of the spin echo is a
in phase (Figure 7-8). If the 180° RF pulse were mirror image of the second half.
at time t, the maximum rephasing would occur
at time 2t. The second half of the spin echo is an FID, and
During imaging, the spin ensemble begins at the first half is a mirror image of an FID.
equilibrium. A 90° RF pulse rotates the net mag-
netization onto the XY plane, that is, the starting The key point that allows rephasing of the
line (see Figure 7-8). The spins precess at the runners is that even though the runners’ speeds
Larmor frequency, but because this is illustrated are different, the speed of each runner is con-
in the rotating frame, the precession is not stant. There is a systematic difference among the
shown. The dephasing results in cancellation of runners, and the effect of this difference can be
the signals at various frequencies, resulting in detected by reversing direction. In the same way,
decreasing signal intensity until there is no signal. the loss of signal due to magnetic field inhomo-
Now a 180° RF pulse is applied, and the result geneities can be recovered because the inhomo-
flips all spins, which then rephase to form a spin geneities are constant. A region of the magnetic
echo with a maximum amplitude at exactly the field that is slightly lower in field strength will
time the spins rephase (runners cross the starting remain so throughout an imaging sequence.
line) (see Figure 7-8). Additional 180° RF pulses However, any random changes cannot be
can be applied to the spin ensemble to produce recovered in this manner. For example, if the
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 81

A Spins in phase B Precessing C Out of phase

D Coherent signal E Less signal F No signal

FIGURE 7-7  Dephasing of spins in the rotating frame. A, In the top row, immediately after the 90° RF pulse, all of the
magnetization is in-phase. B, However, if a portion of the magnetization is in a slightly different magnetic field (dotted
line), it will have a slightly different resonant frequency. C, In the rotating frame, this component of the magnetization
begins precessing and at some point in time will be completely out of phase. D, A more realistic scenario is pictured in
the bottom row, where the magnetization produces a coherent signal. E, Magnetic field inhomogeneities cause some of
the spins to precess in a positive direction, while others precess in a negative direction. F, Eventually all the spins have
cancelled each other out, resulting in total dephasing and no net signal.

3 4
2
4
1
3

B1
1 2

A Signal dephasing B 180° pulse C Rephased signal

FIGURE 7-8  The effect of the 180° refocusing RF pulse is shown in the rotating frame. A, Off-resonance spins precess
in both positive and negative directions. B, The 180° RF pulse (B1) flips the dephased spins to the other side of the
transverse plane. C, The off-resonance spins still precess in the same direction, but now they come together rather than
move apart, forming a spin echo signal.
82 PART II  The Magnetic Resonance Image

TE2
TE

90 180 180 180

RFt

T2
SE#1
FID
SE#2
RFs

T2*

FIGURE 7-9  T2 relaxation time is measured from the envelope of multiple spin echoes.

runners bump into each other momentarily on T2-weighted images are commonly produced
the track, the distance lost is not recoverable by using spin echo and fast spin echo pulse sequences.
reversing directions; the runner still ends up a The 180° refocusing RF pulses eliminate the
little behind or “out of phase” when the runners dephasing of the MRI signal that is due to mac-
“rephase.” roscopic magnetic field inhomogeneities, leaving
In a similar manner, the effects of true T2 only the dephasing due to the microscopic inter-
relaxation are not recovered by the 180° pulses. actions, which are characterized by the trans-
Thus the spin echoes reflect a removal of mag- verse relaxation time (T2).
netic field inhomogeneity but not the removal of To maximize the effect that the T2 will have
true T2 relaxation. Subsequent amplitudes of the on contrast in an MR image, both the TE and
spin echoes are smaller because of true T2 alone, the TR should be set to a long value. If the TE
and this leads to a method to calculate true T2 is long, then those signals coming from spins
(Figure 7-9). with a long T2 will dominate. If the TR is long,
If images are acquired with multiple spin then spins having T1 values that span the whole
echoes acquired at different TE values (Figure range a longitudinal relaxation found in biologi-
7-10), the result is a series of images in which cal tissues will contribute more or less equally to
the change in contrast reflects true T2 relaxation the overall signal.
(see Appendix A). Each time a 180° RF pulse is
used a spin echo results, and each spin echo is
HOW TO MAKE T2*-WEIGHTED
smaller than the previous one and reversed
IMAGES
in polarity.
The time from the 180° RF pulse to the first We have seen how the the timing parameter,
echo is generally equal to the time from the 90° TE, controls the T2-dependence of spin echo
RF pulse to the 180° RF pulse. If multiple and fast spin echo pulse sequences. Similarly,
spin echoes are acquired, the TEs of these echoes TE controls the T2*-dependence of gradient
are often conveniently set as multiples of the echo pulse sequences. The excitation pulse rep-
first echo. etition time, TR, controls the T1-dependence of
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 83

TE15 TE30 TE45

TE60 TE75 TE90

FIGURE 7-10  Panel of six multi-echo spin echo images showing the change in contrast from the most proton density–
weighted (top left image, TE = 15 ms) to most T2 weighted (bottom, right image, TE = 90 ms). All images are from the
same slice in a subject with TR = 2000 ms. (Other parameters: slice thickness = 4.8 mm, FOV = 22 × 18.5 cm, NSA = 1,
matrix = 256 × 162, BW = 200 Hz/pixel, B0 = 3 T.)

spin echo and fast spin echo pulse sequences. In frame, resulting in signal loss. Thus, employing
gradient echo pulse sequences, T1 weighting a gradient echo sequence with a long TE and a
is determined by both TR and the excitation long TR produces a T2*-weighted MR image. In
flip angle. addition, a low flip angle, α, should be used such
T2*-weighted images are produced using that α2≪2⋅TR/T1, in order to minimize the
gradient echo pulse sequences. The absence T1-weighting of the pulse sequence.
of the 180° spin echo refocusing pulse makes T2*-weighted sequences can be used to
the gradient echo sequence much more characterize the age of blood accumulations by
sensitive to magnetic susceptibility–induced field evaluating whether they consist of paramagnetic
inhomogeneities. deoxyhemoglobin, methhemoglobin, or hemo-
Areas of intra-voxel magnetic susceptibility siderin. T2*-weighted imaging is also used to
variation create relatively rapid dephasing of the discriminate cerebral hemorrhages, arteriove-
magnetization in the XY plane of the rotating nous malformations, and cavernomas. All of
84 PART II  The Magnetic Resonance Image

TE3.3 ms TE7.3 ms

TE11.3 ms TE15.3 ms

FIGURE 7-11  Signal losses are prominent in T2*-weighted gradient echo images at interfaces where major magnetic
susceptibility differences occur. These effects become more pronounced as the TE echo delay is increased because the
dephasing process has more time to play out. These four images were acquired at the level of the middle cerebral arteries
(MCA) using spoiled gradient echo pulse sequences with various TE values and TR = 380 ms. Note the decrease of signal
in the temporal region (white arrowhead) and the frontal region (white arrow) due to air–bone interfaces at these sinuses.
Also, oxygenated blood in the MCA produces a susceptibility artifact that presents as a dark line, parallel to the artery
(black arrow). (Other parameters: slice = 4 mm; FOV = 22 cm ×17.8 cm, NSA = 2, flip angle = 25°, matrix = 256 × 208,
BW = 260 Hz/pixel, B0 = 3 T.)

these conditions cause increases in magnetic field but they are also noticeable at tissue/bone inter-
susceptiblity. However, susceptibility-induced faces (Figure 7-11). Susceptibility-induced arti-
artifacts are found in gradient echo images under facts in T2*-weighted gradient echo images
normal conditions and they become more promi- are usually problematic, but T2* weighting
nent with the use of increasing TE values, limit- can provide useful image contrast in some cir-
ing the utility of gradient echo T2*-weighted cumstances, including blood sensitive imaging,
images in many cases. blood oxygen level dependent (BOLD) contrast
These artifacts are particularly severe for functional imaging, and some spine imaging (see
regions in which there are tissue/air interfaces, Chapter 24).
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 85

A B C
FIGURE 7-12  After a 90° RF pulse the longitudinal magnetization has been flipped (A) and it is all in the transverse
plane. If we entirely ignore T2 and T2* dephasing, (B) then the T1ρ processes shrink the magnetization in the transverse
plane while the T1 processes cause the magnetization to grow along the Z-axis (C). Eventually, after about 5⋅T1, the
magnetization is back in equilibrium, totally aligned with the X-axis.

HOW TO MAKE PROTON HOW TO MAKE T1-WEIGHTED

DENSITY–WEIGHTED IMAGES IMAGES
In general, we describe the weighting of RF pulse T1-induced changes in the magnetization are not
sequences based on the relative values of TR measured directly because they represent altera-
and TE. That is, proton density–weighted pulse tions in the magnetization along the Z-axis. Con-
sequences have a long TR in the short TE. sider the state of the magnetization following a
T1-weighted pulse sequences have a short TR 90° RF pulse that is transmitted into the patient
and a short TE. T2-weighted pulse sequences so that the net magnetization is flipped into the
have a long TR and a long TE. Pulse sequences XY plane.
having a short TR in the long TE tend to have The net magnetization seeks to realign
very poor signal-to-noise ratio (SNR) and there- with the external magnetic field. It does this
fore are not used. by shrinking in the −Z direction until it disap-
Typically those pulse sequences that are pears, and then grows in the +Z direction until
deemed proton density–weighted are simply spin it finally relaxes to equilibrium. This process
echo sequences with short TR and short TE. happens because individual proton dipoles
Although some small amount of relaxation has are flipping back to the low-energy state
occurred during these delay times, the amount is (Figure 7-12).
relatively small and the tissue contrast strongly It is important to be aware that the T1 mag-
indicates the relative proton density of the tissues netization recovery process ocurrs concurrently
being imaged. with T2 and T2* dephasing, which cause signal
Note that the shorter the TR is or the greater loss. However, in biological tissues these pro-
the TE is, the poorer the signal-to-noise ratio will cesses ocurr on different time scales (Figure
be. Thus, although proton density–weighted MR 7-13). It is this feature that allows us to separate
images exhibit relatively poor contrast compared out the T1 and T2 image contrast effects in a
to T2-weighted and T1-weighted images, the fairly clean manner and produce T1-weighted
signal-to-noise ratios of these images are uni- and T2-weighted images.
formly superior. This may be a useful property
when anatomical images of great sharpness are
The net magnetization must have a component
desired because the high SNR can often be traded
in the XY plane to detect a signal.
for improved spatial resolution.
86 PART II  The Magnetic Resonance Image

magnetization are usually used in situations where imaging time

is limited (i.e., single slice dynamic imaging,
Logitudinal

Mo
gated cardiac imaging, etc).
1  exp (TR/T1) For T1-weighted gradient echo imaging, short
TR values are used along with large partial flip
A angles (60° to 80°). Again, the specific parame-
100’s of ms
ters used will depend on the B0 field strength of
magnetization

the magnet. This approach allows a great deal of

Transverse

control of the T1-weighted contrast in gradient

Exp (TE/T2*) echo images (Figure 7-14).
Further improvement in the T1-weighting
of MR images can be gained by using an addi-
B 10’s of ms
tional, 180° preparation pulse. Instead of refo-
FIGURE 7-13  The T1 and T2 magnetic resonance relax- cusing the magnetization, like the 180o pulse
ation processes take place on very different time scales.
The T2 dephasing occurs in most biological tissues (with
used with spin echo imaging, this additional 180°
the notable exception of CSF) in a few tens of milliseconds, pulse is placed before the 90° excitation pulse
while T1 processes play out over hundreds to thousands and is used to intentionally invert the longitudi-
of milliseconds. nal magnetization so that it is all aligned along
the −Z axis.
If the 90° excitation RF pulse is applied imme-
T1-weighting of the MR image depends on diately after the 180° inversion pulse, an FID
the history of previous excitation pulses. In con- signal will be produced, which has the same
tructing the image, the intial 90° RF pulse must amplitude but is 180° out-of-phase, compared to
be followed by successive 90° RF pulses corre- the 90° pulse used alone. If a short time is allowed
sponding to subsequent phase endings of the between pulses instead of the 90° excitation
MRI signal. If the intial 90° RF pulse is immedi- pulse being applied immediately after the 180°
ately followed by another 90° RF pulse, the net RF pulse, a T1-weighted signal results as is
magnetization is rotated onto the XY plane and depicted in Figure 7-14.
then again flipped so that all of the magnetiza- The delay time between the 180° and 90° RF
tion lies along the −Z-axis and no signal is pulses is the TI or inversion time (see Chapter 5).
received. Because there is now a little time before the net
If there is a delay between the first 90° excita- magnetization vector is rotated onto the XY
tion RF pulse and the subsequent excitation plane, the net magnetization has a chance to
pulses, which is called the repetition time or TR, shrink somewhat because of T1 relaxation.
then some of the magnetization has a chance to Although still starting out negative, the FID
recover so that there is a useful amount of mag- received is smaller than that which would be
netization flipped into the XY plane and finite obtained if the excitation pulse were to be applied
signals are received following subsequent excita- immediately following the inversion pulse. The
tions. If the TR is of the same order of magnitude initial magnitude of the FID is once again equal
as the T1 values of the tissues being imaged, the to the size of the T1 relaxed net magnetization
image will exhibit contrast that is strongly related vector.
to those T1 values. If this process is continued, with longer and
Generally, T1-weighted images are produced longer TI values, MR signals that can be inter-
using spin echo pulse sequences with TR~500 ms preted for T1 are received (Figure 7-15). In
and TE~20 ms, depending on the B0 of the MRI Figure 7-15, C, the inversion time (TI) is just
system. Gradient echo pulse sequences can also long enough to sample the net magnetization at
be used to obtain T1-weighted images, and these a time when it is zero, switching from −Z to +Z,
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 87

TR  30 ms, NSA  4 TR  100 ms, NSA  2

TR  500 ms, NSA  1 TR  1000 s, NSA  0.5

FIGURE 7-14  This series of spoiled gradient echo images, all having a TE = 15 ms, show how image quality and soft
tissue contrast change as the TR value is increased from 30 ms up to 1000 ms, where the signal averaging was changed
to keep the total scan times approximately equal. The images with short TE are T1-weighted and, as TR increases, the
images become more proton density–weighted. (Other parameters: slice = 4 mm, 22 cm × 17.8 cm, flip angle = 70°,
matrix = 320 × 208.)

so that when the 90° RF pulse is emitted, there This should be no surprise. As was pointed
is no signal. In Figure 7-15, D, TI is long enough out, the initial height of the FID is exactly the
that the net magnetization is back along the same as the length of the net magnetization
+Z-axis, and the FID has a positive initial vector after a delay of TI. Thus the plot is actu-
magnitude. ally the length of the net magnetization vector at
The results of this type of sequential experi- the various times chosen by TI. A value for T1
ment are used to determine T1. A curve like can be obtained by fitting these points to the
that in Figure 6-2 is obtained by plotting the curve given by the equation in Appendix A.
initial magnitude of each FID versus the TI This RF pulse sequence, a 180° pulse followed
(Figure 7-16). by a 90° pulse after delay of TI, is called an
88 PART II  The Magnetic Resonance Image

TI inversion recovery pulse sequence because it

RFt inverts the net magnetization vector and then
allows it to recover before measurement. As
before, the discussion has been simplified with
the assumption that an FID is the measured
RFs MR signal. Actually, spin echoes are measured.
A Examples of inversion recovery images compared
TI to fast spin echoes acquired with the same param-
RFt eters but with the inversion pulse are shown in
Figure 7-17.

RFs T1-WEIGHTED IMAGES VERSUS

B
T2-WEIGHTED IMAGES
TI There are several differences in the techniques
RFt used to make T1-weighted and T2-weighted
images. A variety of T2 weighting values can be
obtained by the use of a multiple echo, spin echo
pulse sequence. T1 weighting is obtained by
RFs manipulating the TR of regular spin echo pulse
C sequences.
TI Inversion recovery images effectively double
RFt the dynamic range of T1 weighting by using
inversion preparation RF pulses followed by a TI
delay, during which T1 relaxation is allowed to
occur. Because of this, the inversion recovery
RFs sequences often require a significantly longer
D time than the spin echo and fast spin echo
FIGURE 7-15  As inversion time (TI) is increased, the mag- sequences used to produce T2 weighting.
nitude of the free induction decay (FID) is reduced by T1 Table 7-1 summarizes the different types of
relaxation until it is zero. With still longer TIs, the FID
amplitude increases as MZ relaxes to M0.
pulse sequences used to manipulate proton
density, T1, and T2 contrast. They are character-
ized by the number of RF pulses used and the
type of MRI signal (spin echo or gradient echo)
they produce. These types of pulse sequences and
this approach to contrast manipulation is funda-
M0 mental to understanding the clinical use of MR
imaging.
Initial T1 relaxation
height d
of c CONTRAST AGENTS IN MRI
FID TI
b In addition to manipulating soft tissue contrast
−M0 a in MR images by adjusting the RF pulse flip
angles and pulse timings, exogenous contrast
FIGURE 7-16  A plot of free induction decay (FID) ampli-
tude versus inversion time can be used to determine T1 agents can be injected into the patient to physi-
relaxation time. The letters of each data point correspond cally change the relaxation times of tissues. Con-
to Figure 7-15. ventional contrast agents, available commercially,
 CHAPTER 7  Manipulating Magnetic Resonance Image Contrast 89

FIGURE 7-17  These images depict how the inversion recovery prepulse can change the contrast of T2-weighted fast spin
echo images to become more T1-weighted. In the top row four images are depicted that were obtained using TEeff =
93 ms, ETL = 16, TR = 7200 ms, matrix = 256 × 232, BW = 285 Hz/pixel, 5 mm slice, 22 cm × 20 cm FOV. The images
shown in the bottom row were acquired with the same parameters except that an inversion pulse with TI = 2250 ms was
used.

TA B L E 7 - 1 The Basic Pulse Sequences Used to Obtain Tissue Contrast Based

on Relaxation Effects of the Various Tissues and Their General Clinical
Applications

Name No. of RF Pulses Contrast Weighting Applications

Gradient echo One T1 or T2* Fast imaging (3DFT)
Spin echo and fast spin echo Two or more PD, T1, or T2 Conventional imaging
Inversion recovery Three T1 or T2 Exclude certain tissues

are principally designed to change the T1 and

T2* relaxation times. CHALLENGE QUESTIONS
A great deal of research is currently underway
to develop new contrast agents for MRI applica- 1. What is the principal reason MRI relaxation
tions in molecular imaging. The range of applica- times cannot be measured directly?
tions of contrast agents in MRI and the potential 2. What pulse sequences can be used to deter-
for new contrast agents in the near future are mine T1 relaxation time?
discussed in greater detail in Chapters 27 3. At what time after the 180° RF refocusing
and 30. pulse is the MR signal most intense?
90 PART II  The Magnetic Resonance Image

4. What is the MR signal used to make an 8. What would result from the following RF
inversion recovery image? pulse sequence: 90° … 180° … 180° …
5. If you conducted a spin echo pulse sequence 180°… ?
in an absolutely uniform B0 magnetic field, 9. Describe the signal obtained during inver-
what would be the results? sion recovery imaging when the inversion
6. During inversion recovery imaging, why is time equals the time that longitudinal relax-
a signal not detected after the initial 180° ation passes through the origin.
RF pulse? 10. Draw the vector diagram that represents
7. Diagram the difference in transverse relax- tissue magnetization after a 90° RF pulse.
ation representing T2 and that representing
T2*.
 CHAPTER

8 
Fourier Transforms in
Magnetic Resonance Imaging

OBJECTIVES
At the completion of this chapter, the student • Discuss how the magnetic resonance (MR)
should be able to do the following: signal is located in the patient (spatial
• Define mathematical transform. localization).
• Describe the use of the Fourier transform in • Define the Nyquist theorem and discuss its use
magnetic resonance imaging (MRI). in MRI.
• Identify the following concepts: spatial • Identify the MRI artifact aliasing and its
domain, frequency domain, and spatial cause.
frequency domain.

OUTLINE
What Is a Transform? Properties of Fourier Flow and the Fourier
What Is the Fourier Transform? Transforms of Image Transform
Why a Transform? Data Sampling and Aliasing
The Frequency Domain Spatial Localization
Too Small to See and the Fourier
Chemistry’s Signature Transform

KEY TERMS
Alias Fourier transform Sampling
Convolution Imaginary part Spatial frequency domain

Jean Baptiste Joseph Fourier (1768-1830) was a The FT has always played an important role
French physicist and mathematician who lived at in digital image processing. Until recently, this
the time of the French Revolution. Among his role was buried in the depths of the derivation
many accomplishments is the derivation of the of the computed tomography (CT) reconstruc-
mathematical transform that carries his name, tion algorithm or in more complex image quality
the Fourier transform (FT). specifications such as the modulation transfer

91
92 PART II  The Magnetic Resonance Image

function (MTF). However, with the appearance numbers is called a function. The relationship
of magnetic resonance imaging (MRI), the FT may be written as follows:
has been called to center stage.
FT is the mathematical mechanism for chang- Transform
ing any of the magnetic resonance (MR) signals
(free induction decay [FID], spin echo [SE], or 2 --f (area) → 4
gradient echo [GRE]) into a nuclear magnetic where the notation f(area) denotes the “area
resonance (NMR) spectrum for chemical analy- function.”
sis or into a diagnostic image. Therefore an
understanding of this transform is necessary, Note that this area function includes the fol-
especially as it relates to MRI. Among other lowing important properties:
features, the FT provides an explanation of
a type of artifact encountered in MR images 1. The function gives a unique result. A particu-
(i.e., aliasing). lar length for the side of a square results in
only one possible value for its area.
2. The function possesses a unique inverse.
WHAT IS A TRANSFORM? Given the area of a square, the length of its
The FT is only one of many transforms in math- side can be computed, and only one answer
ematics. For an understanding of what mathema- is possible. For example,
ticians mean by a transform, perhaps it is best to
provide an analogy. Inverse Transform
In nature, connections between numbers
always occur. For example, the length of the 4 -- f (area -1 ) → 2
side of a square affects the area of the square where the −1 superscript means inverse.
and vice versa. With the measurement of a few
squares, a pattern of values begins to appear 3. General rules regarding this function can be
(Table 8-1). derived. For example, if the length of the side
It is obvious that the area of the square of the square doubles, then the area is multi-
depends on the length of the side of the square. plied by a factor of 4.
A general relationship between these quantities 4. A mathematical formula can be written to
can be defined: to find the area of the square, represent this relationship.
multiply the length of the side of the square by
itself. The area equals the square of the length of Variable Transform
a side. This relationship between two sets of
(area of square) = (length of side)2
or with symbols instead of words:
A = s2
TA B L E 8 - 1 The Relationship Between
the Side and Area 5. The units used are changed by the function.
of a Square For example, if the length of the side is mea-
sured in centimeters (cm), then the area is
Side of Square Area of Square
measured in square centimeters (cm2).
1 1
2 4
4 16 A function establishes a relationship between
8 64 numbers; a transform establishes a relationship
10 100 between functions.
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 93

FT

FIGURE 8-1  With Fourier transformation, a square wave results in a “wavy” pattern.

FT−1

FIGURE 8-2  The inverse Fourier transform (FT −1) of the wavy pattern is a square wave.

oscillating functions whose effects are often seen

WHAT IS THE FOURIER
in the FT. For example, the square function in
TRANSFORM?
Figure 8-1 is sharp edged, yet its FT has waves
The FT is only one of many transforms available; in it. These waves include sines and cosines.
however, its properties make it uniquely useful in The units of the source function and the result-
MRI. The FT can be presented in terms of graphs ing transformed function are different but related.
of functions to see how the FT changes the shape If the source function is a plot of signal intensity
of these graphs. For example, the square wave versus time, then the Fourier transformed func-
function is transformed into the wavy function tion is a plot of signal intensity versus 1/time (i.e.,
by Fourier transformation (Figure 8-1). The frequency).
symbol FT in the figure indicates this mathemati-
cal formulation. The FT of intensity versus time is intensity versus
The FT has properties analogous to the area- frequency.
of-a-square function discussed previously. The
FT gives a unique result; for example, the square
function (or boxcar function) of Figure 8-1 is
WHY A TRANSFORM?
Fourier transformed only into the wavy function
shown. This wavy function is called a sinc func- What is the use of a transform? The answer lies
tion or sin x/x. The amplitude and width of the in the desire to solve problems. Physicists and
square function are related to the amplitude and engineers are always trying to understand the
wavelength of the sinc function. real world and how it reacts to changing situa-
Because the FT is unique, a unique inverse FT tions. They write formulas to represent some part
also exists. Only one function can produce the of the world, and then try to solve these formulas
wavy function of Figure 8-1; that is, the FT can to see how that part will behave. For example,
be undone so that the original function is pro- the response of a bridge to a crosswind can be
duced (Figure 8-2). The symbol FT−1 represents predicted by setting up a set of equations that
the inverse Fourier transform. represent the properties of the bridge and by
A mathematical formula can be written to solving these equations in the presence of a force
define the FT. The exact form of this formula that represents the wind. Unfortunately, these
is unimportant, except that it contains sums equations are often extremely complicated, and
of sines and cosines. Sines and cosines are their solution is not at all obvious (Figure 8-3).
94 PART II  The Magnetic Resonance Image

problem

solution

solution

problem
FIGURE 8-3  Mathematical formulas state problems. The solution often requires a transformation.

real space Fourier space The FT does not really change the information
present in a function. Rather, it represents that
problem FT ÒproblemÓ information in a reorganized way, offering a
new viewpoint on the data. Thus the situation is
viewed in either real space or Fourier space. In
solution FT−1 ÒsolutionÓ
both spaces the same real-world thing is repre-
sented (e.g., an image, an MR signal). The FT
FIGURE 8-4  A magnetic resonance image is obtained by
transforming a signal into Fourier space, reassembling the just offers a new and unique viewpoint on these
data, and computing the inverse transform. data. For example, the MR signal is a function
of intensity versus time (i.e., time domain). The
FT gives a representation of those same data as
intensity versus frequency (i.e., the frequency
Sometimes if a transform like the FT is applied domain).
to source equations, the solution of the trans-
formed equations is easier to obtain than the
THE FREQUENCY DOMAIN
solution of the source equations. The answer to
the FT of the source equation is in Fourier space. Suppose that the real-space function represents
If an inverse transform is applied to the Fourier some sort of signal in time, that is, a representa-
space equation, the result is in real space (Figure tion of the signal intensity as it varies with time,
8-4). This result is the same as if the problem had like an MR signal. The Fourier space representa-
been solved directly. In addition, viewing the tion does not have the same units. The units on
problem in Fourier space sometimes gives unique the horizontal axis in Fourier space are inverse
insights into the situation, insights that are not of the units in real space. In this case, the hori-
obvious in real space. zontal real-space unit is time (e.g., seconds).
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 95

real space Fourier space

(time domain) (frequency domain)

intensity

intensity
FT
time frequency

FT

FT

FIGURE 8-5  The Fourier transformation of a broad Gaussian distribution results in a narrow frequency spectrum and
vice versa.

Therefore the unit in Fourier space is 1/time (e.g., In the complex sound from a symphony, the
1/seconds). ear can distinguish between the high treble pitch
The quantity 1/time (e.g., cycles/second, hertz) of a violin and the deeper bass pitch of a tuba.
occurs often and is frequency. A plot of intensity Indeed, a small percentage of people have abso-
versus frequency is a spectrum. The FT can take lute pitch, the ability to tell the precise pitches
a picture of intensity versus time (i.e., the time (i.e., frequencies) of the sounds they hear. In this
domain) and create a picture of the same signal sense, ears “view” the world in the frequency
represented as intensity versus frequency (i.e., the domain.
frequency domain).
Once again, real-space and Fourier space
views are two different representations of the
Too Small to See
same real-world object. For example, the MR On the left side of Figure 8-5 are three smooth,
signal is a real-space representation of how that bell-shaped (i.e., Gaussian) functions in real
signal varies with time. The FT shows how the space. On the right side are their corresponding
same signal varies with frequency, that is, what Fourier transforms. If these real-space functions
frequencies are present in the signal. represent a signal (i.e., the representation of
The concept of frequency domain or spatial intensity with time), then the Fourier space rep-
frequency domain is easy to recognize. For resentation is a frequency spectrum (i.e., repre-
example, ears do a frequency transformation sentation of the frequencies present).
of the signals (i.e., sounds) that they receive. The time domain representation in the top
The time domain representation of the sound pair of curves shows a wide curve that changes
generated by a complex source such as a sym- slowly over time. The corresponding Fourier
phony orchestra can be represented by a rapidly space representation shows a narrow function in
varying oscillating signal. This signal alone the frequency domain. This means that the signal
conveys little meaning. However, human hearing contains only a narrow range of frequencies. As
takes this time-varying signal and transforms it the time domain signal narrows (i.e., the signal
into frequencies. is made to change faster in time), the curves in
96 PART II  The Magnetic Resonance Image

Real space Fourier space

FT−1

Off
Truncated

FT−1

Roll-off
FIGURE 8-6  If the high frequencies of a signal in Fourier space are chopped off (i.e., truncated), the inverse Fourier
transform (FT −1) results in a ringing appearance at the sharp boundaries in an image.

the frequency domain become broader, indicat- This is unfortunate because no system can
ing that a wider range of frequencies is required. handle an infinite range of frequencies. When
In extreme cases where a sharp signal spike an object is detected with the radio receiver of
exists in the time domain, the range of frequen- an MRI system, the frequencies inherent in the
cies contained in that spike approaches infinity. object must pass through the imaging system. If
Thus the more localized a signal is in time, the the electronics of the system do not pass all the
wider the range of frequencies that must frequencies in the object, then part of the struc-
be handled. In other words, the sharp edges of ture of that object is lost.
objects contain more extremely wide ranges of The system could be designed more carefully,
frequencies and higher frequencies than smooth but there is always a finite limit to the range of
objects. frequencies that it allows to pass. Therefore some
part of the information from the object (i.e., the
A bone–soft tissue interface is a high spatial part of the object contained in the high frequen-
frequency object. cies) is always lost. Various curves show the
effect of this loss (Figure 8-6).
This simple fact has profound implications for From a curve in Fourier space, the view of that
viewing the world. Consider again the square curve in real space can be obtained by applying
wave of Figure 8-1. This sharp-edged object the inverse FT. In the top set of curves, the high
might be an MR signal of a fluid-filled cyst. In frequencies have been abruptly chopped off. This
Figure 8-1, the Fourier space representation is truncation, and the signal is said to be trun-
shows what frequencies are present in the object cated. Truncation results in large oscillations at
as given by the wavy function to the right. Note the sharp edges of the real-space square wave.
that this representation only shows part of the If an MRI system has a sharp cutoff of fre-
Fourier space function; the ripples diminish in quencies, a false, sharp ringing will be detected
height to the right and left but never totally dis- every time the signal changes abruptly. Because
appear until reaching plus and minus infinity. For this ringing is so objectionable, systems are
this object to be truly represented, an infinite designed so that they do not cut off sharply at
range of frequencies must be handled. the edge of their frequency range; rather, they
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 97

Signal Signal
intensity intensity

FT

Time Frequency
FIGURE 8-7  The Fourier transform (FT) of a free induction decay results in a nuclear magnetic resonance spectrum,
which is in the frequency domain.

fade away gradually. A sharp square wave These changes in resonance are changes in
received by such a system would come out with frequency, but the signal received is changing
rounded edges. The information that forms the in time. The FT is the bridge connecting the
sharp edges of the object is lost, thereby causing time domain signal to the frequency domain
the object to become blurred. Spatial resolution representation.
is reduced. The sample FID is the plot of signal intensity
Consequently, there is a limit to the ability to versus time for a complex molecule (Figure 8-7).
image sharp edges and produce fine detail in that If an FT is applied to this signal, a plot of signal
image. This situation can be improved by increas- intensity versus frequency is obtained. The clear
ing the ability of the MRI system to handle high arrangement of sharp peaks of various heights
frequencies. Because there is always a limit to the should be noted because this particular arrange-
frequencies that any MRI system can handle, ment of peaks is the unique chemical signature
there is always a limit to the object size that can for that molecule.
be imaged. A trained NMR chemist learns to recognize
the standard arrangements of these peaks. The
No imaging system can pass an infinite range of relationships of these peaks to one another and
frequencies. their widths and heights indicate the nature of
the bonding between atoms. All of the same
information is contained in the original FID,
although in an obscured form. The FT produces
Chemistry’s Signature a new and useful view of the data.
The MR signal is profoundly affected by the
chemical bonding of the atoms generating the
PROPERTIES OF FOURIER
signal. If a complex molecule emits the MR
TRANSFORMS OF
signal, the signal will have a correspondingly
IMAGE DATA
complex structure.
A nucleus that is bound inside a complex mol- There are a number of properties of the Fourier
ecule will resonate at a slightly different fre- transform which should be understood to facili-
quency than a nucleus in a simple water molecule. tate its use as a tool for image data analysis in
This is due to the magnetic fields of electrons in MRI. We have previously stated and demon-
the atom shielding the nucleus from the B0 field strated in Figures 8-1 and 8-2 that the Fourier
(see Chapter 21). transform and its inverse must always exist and
98 PART II  The Magnetic Resonance Image

FT

 1/
FIGURE 8-8  The Fourier transform of a comb function (left) is another comb function (right) with the vertical lines sepa-
rated by a frequency that is the inverse of the time between the spikes on the comb function. This is an example of the
periodicity of the Fourier transform. The comb function is commonly used to describe the digital sampling process, in
which signals are sampled at discrete points in time.

they must be reversible. This condition is known Fourier transformation process down into its
as orthogonality. most basic components and demonstrate how it
Up to now, only one-dimensional Fourier can be applied to simple images.
transforms have been considered. However, The transformation from the MRI signal to
in imaging we don’t deal with simple one- the MR image itself is actually an inverse Fourier
dimensional Fourier transforms because images transform, so we will look at the transformation
are inherently two-dimensional. Therefore we from the image to the raw data set, which
must use two-dimensional (and sometimes even is perfectly acceptable under the principle of
three-dimensional) Fourier transforms in MRI. orthogonality. The easiest case to consider is an
However, in principle, any multidimensional image that is pure white and has a defined field
Fourier transform can be separated into multiple of view (FOV) of, say x cm × x cm. We only need
one-dimensional Fourier transforms. This prop- to consider one direction at a time (due to the
erty is convienently known as seperability. seperability principle) and we note that this cor-
The third fundamental property of the Fourier responds to a square pulse (also called a boxcar
transform is periodicity, which relates the peri- function) having a width, x.
odic behavior of the FT with its inverse. Simply As we have already observed, the Fourier
put, if the data obtained in one domain (for transform of this pulse will be a sinc function
example, time) consists of discretely sampled with period 1/x (Figure 8-9). Thus the two-
points, then the data in the other domain (for dimensional Fourier transform of a circular cross
example, frequency) will be inherently periodic section of a phantom with uniform intensity will
with data at frequencies that are the inverse of produce a raw data set, described by a sinc func-
the time samples. For instance, MRI data are tion, that has a peak in the center (i.e., zero
sampled at discrete points in time by the analog- frequency) with an amplitude that is set by the
digital converter electronics so that the MR average signal intensity of the image, and peaks
image data are periodic and consist of duplica- of decreasing amplitudes along the X-axis and
tions of the original signals, which may appear Y-axis at increments of 1/x away from the center
as image artifacts. of the frequency space (Figure 8-10).
Periodic sampling of the raw MRI data is rep- Let’s now consider a slightly more compli-
resented mathematically by a comb function, cated image with alternating black and white
which is a series of stick functions that are horizontal transitions that are equally spaced
applied at regular intervals, τ. The Fourier trans- x/10 apart. We need to understand how the
form of the comb function produces another boxcar function, which defines the FOV, inter-
comb function, but with frequency spacings of acts with the sinusoidal wave function that
1/τ (Figure 8-8). describes the successive bright and dark regions.
In the following examples, we will use the Mathematicians call this interaction a folding
three principles described above to break the or convolution of the two functions. That is,
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 99

T2s also a peak in the negative frequency range at

−10/x.
The resulting Fourier tranform is the folding
of the FOV with the sinusoidal waveform, which
is presented as three sinc functions centered
at the −10/x, 0, and +10/x frequencies (Figure
Time
8-11). In this example, we see how the Fourier
transform has been used to describe the proper-
A 0
ties of a simple image as a representation of
frequencies inherent in the spatial characteristics
of the original image, which we call the spatial
frequencies.
0.25 Hz
Now let’s look at this simple image we have
Complex created and see how changing it affects the
FT Fourier transform of the image. First, as you
Frequency might expect, if we change the sinusoidal inten-
sity alterations from horizontal to vertical, the
B 0.5 Hz
Fourier transform will be three sinc functions
that also align in the vertical direction in the
spatial frequency domain (Figure 8-12).
If the alternating patterns run along the diago-
Magnitude nal, so will the three sinc functions in the Fourier
FT
transform. If the bands of alternating image
intensity are spread further apart, this will
produce lower spatial frequencies and the three
C 0 Frequency sinc functions will move closer to the center of
FIGURE 8-9  The Fourier transform of a rectangular func- the spatial frequency domain. Conversely, making
tion (A) is a sinc function (sin x/x). As the duration of the the bands of alternating image intensity nar-
rectangular pulse is increased, the frequency difference rower will produce higher spatial frequencies
between zero-crossings of the sinc function will decrease.
and the three sinc functions will spread out away
B, A complex Fourier transform will produce both negative
and positive values. However, negative values are not easy from the center of the spatial frequency domain
to represent in an image, so we typically use the magni- (Figure 8-13).
tude FT for MRI. This produces a variant that is the abso- In MR imaging the raw data is obtained in the
lute values of the sinc function (C). spatial frequency domain and an inverse two-
dimensional Fourier transform (2DFT−1) is used
to produce the image. Of course most MRI
images are more complicated than the simple
they are folded together so that the resulting sine frequency patterns we have just discussed.
function takes on properties of the two original However, the same principles apply and can
functions. be seen in the raw data sets acquired from
In this case the fundamental sinc function, MRI phantoms that contain regular geometric
which is the Fourier transform of the entire FOV, patterns.
is convolved with the Fourier transform of the Two images from the American College of
sine wave function that contains a peak at the Radiology’s (ACR) MRI accreditation phantom
zero frequency, a peak at the frequency 10/x, and and their associated raw data sets illustrate this
(in order for the mathematics of the Fourier principle. In Figure 8-14 the raw data pattern,
transform to take a form that is reliable to use) dominated by a two-dimensional sinc function,
100 PART II  The Magnetic Resonance Image

FT

Rectangular profile
FT Sinc profile

FIGURE 8-10  The two-dimensional version of the sinc function is shown as the magnitude Fourier transform of an
image of a circle of uniform intensity.

FT

Sinusoidal wave
3 sinc functions
FT
FIGURE 8-11  The sinusoidal function in the horizontal direction produces three peaks, one at the central frequency and
one each at the positive and negative values of the spatial frequency of the sinusoidal pattern. In the magnitude Fourier
transform, this pattern is convolved with the underlying sinc function, which is the Fourier representation of the entire
imaging space, to produce three sinc functions.
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 101

FT

FT

FIGURE 8-12  Rotating the image in Figure 8-12 also rotates the sinc functions in the magnitude Fourier transform. The
rotation of the Fourier transform that occurs with a rotation of the image is an important property of Fourier
transforms.

is shown on the left, and this we already know intervals in both the vertical and horizontal
is the Fourier transform of the boxcar function. directions. Upon inverse Fourier transformation
The 2DFT−1 reveals an image of the flood section we obtain a grid pattern image produced from
from slice #7 of the ACR phantom, shown on slice #5 of the ACR phantom. We now under-
the right. stand that the grid pattern was produced from
In Figure 8-15 the raw data set shows that the the smaller sinc functions spaced at frequencies
frequency components have been convolved, that are inversely proportional to the distances
with a strong central sinc function appearing at separating the horizontal and vertical dark lines
the zero of the spatial frequency (at the center), in the grid pattern.
which is due to everything within the diameter The information contained in an image is the
of the phantom, while peaks appear at regular same as that contained in its two-dimensional
102 PART II  The Magnetic Resonance Image

FT

Sinusoidal wave
3 sinc functions
FT
FIGURE 8-13  If the image contains a sinusoidal pattern in which the positive and negative oscillations are three times
narrower, then the magnitude Fourier transform representation will be three sinc functions that have three times the
frequency separation.

FT1

Sinc profile Rectangular profile

FIGURE 8-14  The raw data obtained from slice #7, the flood section of the ACR MRI accreditation phantom (shown
on the left) is subjected to a magnitude inverse Fourier transform to produce the MR image of the phantom (right).
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 103

Raw data FT1 Image data

FIGURE 8-15  The raw data obtained from slice #5, the geometric distortion section of the ACR MRI accreditation
phantom (shown on the left), are subjected to a magnitude inverse Fourier transform to produce the MR image of the
grid insert (right). The underlying sinc function is evident, as are the peaks that appear at +/− values of the spatial fre-
quency of the grid pattern in both the horizontal and vertical directions.

SPATIAL LOCALIZATION AND

Fourier transform. Likewise, the information
THE FOURIER TRANSFORM
contained in the raw MRI data set is the same as
that contained in its two-dimensional inverse For an MR image to be made, the origin in space
Fourier transform, which is also the processed for each part of the signal must be known. Unfor-
image. In one case the data is expressed as signal tunately, only one signal at a time is received
intensity as a function of spatial position, and in from the patient. Therefore spatial information
the other it is expressed as signal intensity as a must be encoded into each such signal (see Chap-
function of spatial freqeuncy. ters 13 and 14).
The preceding examples have been used to In a uniform external magnetic field, all nuclei
give you a feel for how the Fourier transform resonate at the same frequency, called the reso-
process works. It is a remarkable thing that using nant frequency. If a gradient magnetic field that
digital computational methods, the Fourier varies uniformly in the Z direction is added to
transformation can take the very complex infor- this primary magnetic field, the spins on the −Z-
mation about the spatial frequencies of the signal axis (relative to the isocenter of the magnet) will
obtained from the cross section of a patient’s be in a lower magnetic field than those on the
head and produce a clear image of the cranial +Z-axis. Therefore they will resonate at a lower
anatomy. frequency. This difference in frequency directly
104 PART II  The Magnetic Resonance Image

glob A
+
A FT glob B
B

frequency domain
without gradient

glob A glob B
FT
A
B
frequency domain
with Z gradient
BZ

FIGURE 8-16  When a Z gradient magnetic field is applied, the same tissue results in different peaks in the frequency
domain.

relates to the position of the spins along the approach can also be taken. Such a method
Z-axis and the amplitude of the gradient mag- would energize only one part of the object so that
netic field. any signal received would come only from that
Two globs of fat are shown at different posi- part, rather than from the entire object.
tions in the Z direction (Figure 8-16). With no Z This method uses the same gradient magnetic
gradient magnetic field, the two globs resonate field system as seen in the previous methods. For
at the same frequency and contribute to a single a signal to be received from the spins in only one
peak in the frequency domain. When a gradient object, the initial RF pulse must excite only those
magnetic field, GZ, is added, this single peak spins. For example, if the spins in object B are
splits into two peaks, one peak for each glob, energized and the spins in object A are left undis-
each of which is now at a different frequency. turbed, any signal received would come from
The frequency difference between the peaks object B alone.
directly relates to the distance between the globs For the precessing nuclei to absorb energy, the
of fat in the Z direction. The stronger the gradi- RF signal must exactly match the frequency of
ent magnetic field, the further apart the peaks in precession of the spins. Because objects A and
the frequency domain; thus it is easier to separate B are in a gradient magnetic field, the spins
objects in space. in the objects have slightly different resonant
frequencies. Therefore an RF pulse with a fre-
Gradient magnetic fields provide spatial localiza- quency distribution unique to object B, not that
tion of the MR signal. of object A, is required if object B is to be imaged
(Figure 8-17).
This process relies on the FT. Because an MR For the RF signal to be actually generated,
signal is a variation of intensity as a function of however, knowledge of the signal as a function
time, an FT must be applied to the signal to view of time is required. The FT allows a change
the frequency distribution that is related to the between time and frequency. To go from the fre-
spatial distribution. quency domain to the time domain, one must
In the example in Figure 8-16, each fat glob apply the inverse FT. This exercise provides the
generated a signal. A gradient magnetic field shape of the RF pulse that must be used. If this
was used to determine where along the Z direc- shaped RF pulse were transmitted into a patient,
tion the signal from each originated. Another only the spins in a chosen part of the patient
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 105

FT−1

A
B
time

BSS

FIGURE 8-17  A radiofrequency pulse containing only the frequencies of tissue B is obtained from the inverse Fourier
transform (FT −1) of the frequency distribution.

FT

real part imaginary part

FIGURE 8-18  The Fourier transform (FT) has two parts, real and imaginary.

would be excited. This isolates the MR signal in One general property of the FT is that if the
a narrow section of the patient. real-space function is an even function, then the
In these simple examples, spatial information FT of that function has an imaginary part that is
has been encoded in only one dimension. Similar zero, and the nonzero information is contained
methods, which are also heavily based on the use in the real part. An even function is one that is
of the FT, are used to obtain spatial information symmetrical around the vertical axis (i.e., the
in all three dimensions. right and left halves of the function are mirror
images). The simple functions considered to this
point (i.e., the square wave function and the bell-
FLOW AND THE FOURIER
shaped Gaussian functions) have been even func-
TRANSFORM
tions and therefore have had an FT with an
A true FT generates two parts of the square wave imaginary part equal to zero.
function (Figure 8-18). These parts are called the In most cases the SE from an MRI system is
real part and the imaginary part. These names also an even function. As the spins come back
are imaginative names traditionally used by into phase, the signal intensifies to its maximum
mathematicians for these mathematical parts. and then relaxes back to zero as the spins again
Therefore the imaginary part is just as real as the dephase. The signal produced (i.e., the SE) is
real part. They could have been called anything: symmetrical because the process of rephasing
parts A and B, left and right, or Brenda and Fred, and dephasing is symmetrical.
for that matter. Theoretically, an image produced from a set
Until now, the imaginary part of the transform of SEs actually results in two images: a real part
has been ignored, because this part has been zero image and an imaginary part image. However,
for the examples used. However, under certain because the SE is symmetrical, the imaginary
circumstances, the imaginary part of the trans- part image contains noise but little useful
form contains some useful information. information. As a result, most MRI systems use
106 PART II  The Magnetic Resonance Image

magnitude reconstruction in which both the real computers. This special optimized form is called
and imaginary moduli are used. the fast Fourier transform (FFT).
In some cases, however, the SE is not sym- When an RF signal is sampled, how much of
metrical, as when there is motion, for example, the signal should be measured and stored? How
blood flowing in a vein. If the blood moves in rapidly must the signal be sampled to give a good
the direction of one of the gradient magnetic representation of it? If rapidly sampled, the data
fields, with time it experiences a different mag- points are spaced extremely close to one another.
netic field as it moves from one position in the This always provides a good representation of
gradient magnetic field to another. the signal.
Thus when an SE is received from moving Fast signal sampling creates practical prob-
blood, the rising shape of the SE is not the same lems. It is more difficult to design electronics to
as the falling shape, and the SE can no longer be sample data quickly; therefore the equipment
exactly an even function. Therefore when the becomes more expensive and error prone. The
image is formed, the imaginary part of the image more data sampled, the more data there are to
is not zero, and the real part has missing or dis- store. When there are more data to store, more
torted information. For simple imaging pulse computer memory is required, thereby making
sequences, this motion can generate artifacts in the process more expensive. As more data are
the image. collected, it takes more time to analyze them and
In practice there are sophisticated pulse longer to reconstruct the images.
sequences designed to suppress these motion arti- On the other hand, if too little data are
facts. On the other hand, the detection of moving sampled, important parts of the signal may be
material, especially blood flow, is of special inter- missed, and important information may be lost.
est. Special pulse sequences designed to enhance Therefore it is important to know whether there
and quantify this effect are continually under is an optimum rate at which to sample the data.
development (see Chapter 23). The optimum rate is that which provides an
adequate representation of the data. Oversam-
pling is common and produces small signal-to-
SAMPLING AND ALIASING noise gain. The FT can help answer this question
In all MRI systems, the FT is performed by a and explain what happens if too few points in
computer. A computer does not deal with con- the data are sampled.
tinuous curves like the graphs of signals presented When a real-space function is Fourier trans-
earlier; rather, it manipulates individual (i.e., dis- formed, only a finite range of frequencies are
crete) numbers. This has several important con- needed to represent this signal (i.e., frequencies
sequences. The data of the continuous MR signal up to some maximum value fmax) (Figure 8-19).
must somehow be changed to individual numbers. The required range of frequencies is from −fmax
This change is done by a process called digi- to +fmax. Therefore the width of the frequency
tization or sampling. The intensity of the signal band is 2fmax and is symbolized as Δf. The elec-
is sampled, measured, and stored at regular inter- tronics are then designed to handle only the range
vals. The results are a series of data points that of frequencies up to and including this maximum
give a representation of the original continuous frequency (i.e., up to and including fmax Hz). Any
signal when they are connected. frequency above this limit cannot pass through
A form of the FT that handles discrete numbers the system, but because no such frequencies exist
rather than continuous curves must be available. in the desired signal, this is irrelevant.
This form of the FT is called the discrete Fourier This real-space curve is sampled with data
transform. Because FTs are commonly imple- points spaced at some fixed interval and sent
mented on computers, the discrete FT has been through the system. Is the result of using the
optimized for the binary architecture of such sampled data the same as if the continuous signal
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 107

Time domain Frequency domain

FT

time − fmax fmax +

frequency

FT

1/∆t
FIGURE 8-19  The Fourier transform (FT) of the discrete function of the sampled signal results in clones of continuous
functions.

were Fourier transformed? The FT of the discrete always be greater than the frequency width of
sampled function is a continuous function, not the curve itself (−fmax). This is mathematically
discrete as may be suspected (see Figure 8-19). expressed as follows:
The difference between the FT of the continuous
and discrete functions is that the FT of the Sampling Theory
discrete function has additional clones of the
desired FT. 1/ Dt > -fmax
There are an infinite number of these clones
of the desired FT spaced evenly up and down The wavelength is 1/frequency and 1/maximum
the frequency spectrum and extending to both frequency is 1/minimum wavelength. Sampling
plus and minus infinity. These clones do not theory can be rearranged to become the Nyquist
matter. Because the system only passes frequen- sampling theorem.
cies up to fmax Hz, all the clones are cut off,
provided they lie at frequencies beyond fmax Hz Nyquist Theorem
(i.e., the maximum frequency handled by the
system). Dt < l min/2
This last condition is critical to the proper
representation of the data. The clones are spaced Therefore at least two points must be sampled
apart in the frequency domain by 1/Δt. As the within the smallest wavelength within an object.
sample points are moved farther apart in real
space, the clone curves in frequency space creep
For undersampling and the resulting aliasing to
closer to the central curve. Trouble occurs when
be avoided, more than two samples must be
the bottom point of the first clone curve begins
taken each cycle.
to touch the top point of the central curve. The
sample points can be widened until this condi-
tion occurs and no farther. The wavelength of a simple oscillating signal
More precisely, the sampling distance between is easy to determine (Figure 8-20); it is the dis-
the clone curve and the central curve (1/Δt) must tance between crests or valleys or any two similar
108 PART II  The Magnetic Resonance Image

correctly sampled correct reconstruction

incorrectly sampled erroneous reconstruction

FIGURE 8-20  If the sampling frequency is too low, the reconstruction will be false and result in aliasing.

FIGURE 8-21  Signals from both the ante-

rior and posterior of the head fall outside
the prescribed FOV (field of view). The
signals are thus undersampled and recon-
structed as originating within the FOV but
from opposite sides. (Courtesy Bill Faulkner,
Chatanooga, TN.)

points of the oscillation. According to the Nyquist by connecting the points is possible (see Figure
theorem, a sample must be taken at least twice 8-20). The reconstructed curve matches the origi-
in that wavelength. nal signal through connection of the dots. If the
If more than two samples are taken in each Nyquist theorem is purposely disobeyed so that
wavelength, an attempt to redraw the curve fewer than two points are sampled in every
 CHAPTER 8  Fourier Transforms in Magnetic Resonance Imaging 109

wavelength, a false reconstruction of the signal

CHALLENGE QUESTIONS
results (see Figure 8-20).
The reconstructed signal has a longer wave- 1. The FT changes the MR signal from inten-
length. Not only does the original signal match sity versus time to what quantity?
these sampled points, but it also matches a longer 2. What artifact results when the MR signal is
wavelength signal. This second wavelength (i.e., not adequately sampled?
frequency) signal is an alias. 3. What is another name for spatial frequency
domain?
If the frequencies in a signal are undersampled, 4. What is the minimum sampling rate of an
bogus frequencies result in the output and an MR signal that will ensure aliasing does not
aliasing artifact results. occur?
5. Whose name is applied to sampling theory?
Because MR images use frequencies to encode 6. What property of the image will be affected
position information, aliasing usually results in if the space between the sampled raw data
the image appearing to be “wrapped around” lines is doubled?
another part of the image. For example, a portion 7. Identify three examples of tissue with high
of the left part of the image appears as a ghost spatial frequency components.
on the right part of that image and vice versa 8. Are there any other disadvantages to over­
(Figure 8-21). sampling an MR signal?
9. Which component of the MRI system is
principally responsible for locating the posi-
tion of the MR signal in the body?
10. Graphically, show how the FT of a square
wave and an SE should appear.
 PART
III
The Imaging System
 CHAPTER

9 
Magnetic Resonance
Imaging Hardware

OBJECTIVES
At the completion of this chapter, the student • Discuss the purpose of shim coils.
should be able to do the following: • Identify the principal controls on the MRI
• Name the three major components of a operating console.
magnetic resonance imaging (MRI) system and • Describe distinguishing features of the MRI
the subassemblies of each. computer.
• List the three types of MRI systems and
describe features of each.

OUTLINE
The Gantry Image Processing/Display/ Magnet Power Supply
Superconducting Magnetic Manipulation Sequencing System
Resonance Imaging System Magnetic Resonance Imaging Digital Signal Acquisition
Resistive Electromagnet Computers System
Imaging System Types and Functions of Digital Image Processing
Permanent Magnet Imaging MRI System Computers System
System Storage Capacity Ancillary Equipment for the
The Operators’ Console Computer Speed MRI Suite
Start-Up MRI System Electronics
Image Acquisition Frequency Synthesizer
Prescan Calibrations Radio Frequency Amplifier
Image Scanning Gradient Coil Power Supply

KEY TERMS
Field of view Gantry Lorentz force
Frequency synthesizer Isocenter Sequencing system

111
112 PART III  The Imaging System

The basic physics of magnetic resonance imaging The gantry contains the main magnet and
(MRI) have been covered in the previous chap- several other electromagnetic devices essential to
ters, and the equipment used in the process producing MR images. With the exception of
is discussed in this and the following two the table, there are no moving parts in the
chapters. MRI gantry. The user interface of the operating
An MRI system contains three major compo- console often resembles the manufacturer’s CT
nents, each of which is usually in a separate and nuclear medicine consoles, and although
room and consists of several subsystems. The many of the control designations are similar, they
major components are the gantry, the operating also serve different functions.
console, and the system electronics cabinets. The MRI system electronics include a
With the covers on, an MRI system resembles powerful and fast signal acquisition computer,
a computed tomography (CT) imaging or a radio frequency amplifiers, gradient power sup-
positron emission tomography (PET) system. plies, frequency synthesizers, a pulse sequencing
However, when you take the covers off, most system, and digital image processing systems.
of the similarities end. The principal compo- The system electronics are often located in an
nents of an MRI system are shown in a block adjacent, air conditioned room.
diagram in Figure 9-1. Each of the three types of MRI magnet
systems—superconducting electromagnet, resis-
tive electromagnet, and permanent magnet—uses
Magnet, operating console, and system electron-
similar computers, and the operating consoles
ics are the three principal components of an MRI
have similar functional controls that appear the
system.
same. However, the gantries are completely

System
electronics

Operating
console Gantry

FIGURE 9-1  The principal components of a magnetic resonance imaging system are the magnet gantry, which is located
in the magnet room; the operating console, which is in the operator’s room; and the system electronics, which are in
the electronics room. This diagram focuses on the various electronics components comprising the system electronics,
including the pulse sequence control, gradient and RF power, and signal processing subsystems.
 CHAPTER 9  Magnetic Resonance Imaging Hardware 113

different and each has a distinctive appearance.

Because most MRI systems in use today are
superconducting, the discussion in this chapter
focuses more on that specific type.

THE GANTRY
The gantry can be intimidating, especially after
the patient is placed on the examination couch
and moved into the patient aperture. The patient
then hears the resounding thump, thump, thump
of the gradient coils, suggesting that this is indeed
a big and intimidating machine. The MRI system A
is not a machine that has moving parts; rather it
is an imaging system with no moving parts. 15
However, the gantry does have many subsystems
and several different electromagnetic coils. 12 0.5 Tesla

Radial distance (m)

1.5 Tesla
3.0 Tesla
Superconducting Magnetic 9
Resonance Imaging System
Figure 9-2 shows a typical superconducting MRI 6
magnet in cross section. Superconducting MRI
magnets are approximately 3 m across by 3 m 3
high, with a length that ranges from 1.4 to 1.6 m.
The massive size is due principally to the need to
have the primary magnetic coils at a super-cooled 3 6 9 12 15 18
temperature or cryogenic state. This cryogenic B Axial distance (m)
state is maintained by putting both the coils and FIGURE 9-2  A, This photograph shows a 1.5 T supercon-
the cryogens in insulating chambers. ducting magnet with the covers off. B, This cross-sectional
diagram depicts an idealized, six-coil superconducting
Cryogens are liquefied gases that produce super- magnet. The superconducting coils are located inside a
dewar, or thermally isolated vessel, containing liquid
cold temperatures near absolute zero. Liquid helium (lHe), which is shaded. The 20º K reflective shield
helium is the cryogen that is used primarily to and vacuum thermally uncouple the container of lHe from
cool MRI magnet coils. the environment. The six short, main magnet supercon-
ducting solenoid coils are immersed in the lHe and, once
The gantry of a superconducting MRI system powered up, sustain current indefinitely as long as they
are maintained at a low temperature. Two active-shielded
can be considered to have three subassemblies: coils are designed to reduce the extent of the fringe mag-
the patient couch, the primary electromagnet netic fields into the scanner room. (Active shielding in the
assembly, and the various secondary electromag- magnet can reduce the magnetic fringe field distance by
nets (Figure 9-3). The secondary electromagnets, about one-third on a 1.5 T system and by almost one-half
used for generating the pulsed gradient and RF on a 3.0 T magnet.) The different number of lines on each
coil suggest different current densities. The service turret
magnetic fields, are at room temperature, unlike at the top is where the superconducting leads are attached
the primary electromagnet, which is immersed in to power up the magnet and where the thermally insu-
liquid helium. lated tubing is inserted for filling the magnet with
The patient couch performs the two functions cryogens.
of support and position. The couch should, at a
114 PART III  The Imaging System

Primary
magnets

Localizing
laser

Zero couch
position
Secondary
Localizing
magnets
laser
FIGURE 9-3  The patient couch, the primary electromag- FIGURE 9-4  Two or three positioning laser lights are
net assembly, and the secondary electromagnets are the
adjusted to intersect on the axis of the primary magnetic
three superconducting magnet subassemblies.
field of a magnetic resonance imaging system.

minimum, be able to accept patients who weigh

TABLE 9-1 Minimum Specifications
up to 150 kg (~330 lb) at near floor level and
for an MRI Patient-
raise the patient with a power assist to the level Positioning Couch
of the gantry patient aperture. From this position
outside the gantry, the couch should be capable Descriptor Performance Standard
of moving to the imaging position under power
Patient capacity 130 kg
assist to within ±1 mm. Lift speed 1 cm/s
Precise positioning is essential during exami- Translation speed 10 cm/s
nation. The imaging volume in superconducting Position accuracy ±1 mm
magnets is approximately a 40 cm diameter
sphere. In order to image extended sections of MRI, Magnetic resonance imaging.
the body, such as a full spine, the couch must
move in coordination with the scanning protocol cryostat. The innermost chamber of the cryostat
to position the patient in the sweet spot, which houses an aluminum cylinder around which the
is called the isocenter of the magnet. superconducting wire is wound (Figure 9-5).
This precision is obtained with specialty gears
and electronic registers. Each revolution of the The cryostat is a large insulating container in
smallest drive gear corresponds to a 1 mm move- which liquid helium and the superconducting
ment of the couch. Each revolution also activates magnet coils reside.
an electronic counter so that the couch position
can be visually displayed. Table 9-1 lists minimum Similarly, the secondary magnetic coils are not
acceptable specifications for the patient couch. visible. They are also covered by a thick plastic
At installation, the service engineer adjusts the liner inside the magnet bore. The relative posi-
two or three positioning laser lights to intersect tion of these magnetic coils in the gantry is shown
at a point on the axis of the MRI gantry (Figure in Figure 9-6.
9-4). Usually, this position is then set at zero on There are three sets of secondary magnetic
the couch position indicator. coils, and they are independent. Nearest to the
The primary electromagnet assembly is not patient is the radiofrequency (RF) coil. The RF
visible. It is enclosed within a decorative plastic coil produces a magnetic field that oscillates at
housing. Even with the housing removed, the the resonant frequency of the hydrogen nucleus.
primary electromagnet assembly cannot be seen The RF coil is designed to produce a magnetic
because it is deep within the chambers of the field that is perpendicular to the primary
 CHAPTER 9  Magnetic Resonance Imaging Hardware 115

Primary coils
in a cryostat

Patient
aperture

Shim Gradient RF coil

coils coils
FIGURE 9-7  Relative position of the shim coils, gradient
coils, and radiofrequency (RF) coil.
FIGURE 9-5  Relative position of the primary magnetic
coils in the magnet.
Between the gradient coils and the primary
electromagnet assembly for earlier systems, shim
Secondary
coils were positioned to make the B0 field
electromagnet
coils more homogeneous (uniform field intensity).
Patient Usually the shim coils were at room temperature,
aperture but in advanced MRI systems, they were in the
cryostat.
Currently, shim coils are typically found only
in MRI systems operating at 3 tesla or higher
because of their expense and the desire to save
the additional bore space that they occupy. In
MRI systems at lower field strengths, long bars
FIGURE 9-6  Relative position of the secondary magnetic
coils in the magnet. of ferromagnetic metals (called shim stock) are
placed in trays along the inside of the magnet and
used with gradient magnetic field offsets to
magnetic field. This RF coil, often referred to as obtain excellent B0 homogeneity.
the “integrated body coil,” is a separate, remov-
able assembly and therefore is not cryogenic but
Resistive Electromagnet
remains in place during all system operations and
Imaging System
is kept at room temperature (Figure 9-7).
The secondary magnetic coils located closest The patient couch of a resistive electromagnet
to the patient aperture are the gradient coils, imaging system appears the same as that of a
which are also at room temperature. These coils superconducting MRI system and will have
are large electrical conductors that produce the similar performance characteristics. The coils are
pulsed gradient magnetic fields. They are switched wound in four to six rings, each containing a coil
on and off rapidly. This current switching results of wire conducting a large electrical current, that
in the conductor being subjected to a force, called ranges from 30 to 50 amperes (A) depending on
the Lorentz force, that is at right angles to both the size of the magnet and B0 field strength
the direction of current flow and the primary required. The two large coils produce the B0
magnetic field, and this force causes the thump- magnetic field, whereas the two smaller coils on
ing sound that is commonly heard during scan- each end help to extend the length of the field
ning. This is the same prinicple under which and make it uniform.
loudspeakers operate, so we may think of the The secondary electromagnets of a resistive
MRI system as a low-fidelity loudspeaker. magnet imaging system are often wound in a
116 PART III  The Imaging System

B0

Adjusting screw

Brick assembly
Pole face

Iron yoke

FIGURE 9-8  A permanent magnet MRI system.

flat plate that is put onto the pole caps of the There are no shim coils in a permanent magnet
magnet. This subassembly defines the patient MRI system. Shimming the primary magnetic
aperture. Typically, the RF coil system is a sepa- field is accomplished by mechanically adjusting
rate operator-interchangeable assembly. the two finely machined pole pieces.
Resistive electromagnet MRI systems are The primary advantage of both permanent
rapidly disappearing from the scene. A few and resistive MRI magnets is that they are
remaining 0.2-T and 0.5-T magnets may be resis- much more open than superconducting imaging
tive with an iron core. These magnets look like systems. These systems allow greater accommo-
large C-arms and produce a vertical (floor-to- dation for large patients or patients who may
ceiling) B0 field. experience claustrophobia during the study,
and also allow off-axis body parts such as the
shoulders to be moved to the center of the B0
Permanent Magnet Imaging System magnetic field. However, these systems are grad-
The subassemblies of a permanent magnet ually being replaced by midfield (0.7 T to 1.2 T)
imaging system are not visible because of a deco- superconducting magnet systems, in which two
rative housing. There is no hint of how the subas- coaxial superconducting coils are used to provide
semblies appear except for the RF coils, which a large gap in which patients can be scanned
are identifiable and operator interchangeable. (Figure 9-9).
The cutaway view of the permanent magnet
MRI gantry in Figure 9-8 shows that the primary
THE OPERATORS’ CONSOLE
magnetic field is produced by two assemblies of
bricklike magnets. These primary magnets are The operating console of an MRI system gener-
attached to a massive iron yoke. ally has two sets of controls that are found on
The iron yoke plays the same role with the the operating console of an MRI system. One set
MRI gantry as it does for a transformer and is is for image acquisition and the other for image
similar in design. The iron yoke provides a return processing. Some controls are activated by special
path for the lines of the primary magnetic field. function keys, but most are under computer
The yoke’s presence increases the B0 magnetic command. The operator typically responds to a
field intensity within the patient aperture. video prompt by keying commands using a
 CHAPTER 9  Magnetic Resonance Imaging Hardware 117

FIGURE 9-9  A magnetic resonance imaging operating console. (Courtesy Hitachi Medical of America.)

mouse or touchscreen interface. Independent 2. Emergency off—This should be used only if

workstations for physicians’ reading rooms, the patient is in imminent danger or if contin-
which have advanced processing capabilities, are ued operation could result in damage to the
also available as an option on an MRI system. imaging system.
The image acquisition controls for MRI 3. Intercom—This device allows communication
systems typically have many pages of parameters with the patient. Often, especially when the
to select. The principal control functions on the operator cannot directly see the magnet bore
console of an MRI system are discussed in the through a window, a closed-circuit TV system
following sections. is installed to allow continuous visual moni-
toring of the patient in the magnet.
4. Scan control module—The pulse sequence
Start-Up prescription allows the operator to select the
1. Power on/off—This is usually a key switch or pulse sequence, pulse and delay timings, and
push switch that is used to energize the system. other important parameters for the scan.
118 PART III  The Imaging System

5. Image-guided graphical prescription—Allows scan time, thereby decreasing the number of

placement of slices, the number of slices, and slices per scan available with multislice pulse
the gaps between them, displayed graphically sequence techniques. The longer the spin
on a set of pilot (or scout) images. echo time (TE), the more T2-weighted (T2W)
6. Image display system—Allows the operator to will be the image. The inversion time (TI) is
review the images just acquired, using display the time between the 180° and 90° RF pulse
windowing functions. Additional features in an inversion recovery pulse sequence. TI
include parametric imaging processing (i.e., also controls T1W contrast, but having a TI
diffusion images) and the ability to perform increases overall scan time.
measurements using electronic calipers and 4. RF pulse flip angle selection—For gradient
region-of-interest statistics. echo imaging, the excitation RF flip angle
(FA) is an important parameter for optimiz-
ing image contrast and SNR while minimiz-
Image Acquisition ing image artifacts. In spin echo and fast spin
1. Tuning controls—Several mouse clicks under echo imaging (especially at 3 T and higher
computer command are designed to adjust B0) timing parameters may be constrained by
the resonant frequency of the system to the potential for RF heating of the patient as
accommodate the patient or body part being characterized by the specific absorption rate
imaged. More often, this is an automatic (SAR). Reducing the FAs of refocusing RF
function. The following image parameters pulses from 180° to 120° will reduce spin
can be input before each image acquisition; echo signal amplitude but also reduce RF
however, most clinics have a set of standard heating since SAR ∝ FA2.
protocols that have been developed by the 5. Matrix size—The number of pixels is usually
radiologists and the CT technologists for 256 × 256. A larger matrix results in better
particular types of studies. Thus only a few spatial resolution and poorer signal-to-noise
parameters, such as field of view and image ratio (SNR), but also requires a longer
orientation, are modified before each study imaging time. Pixels can be made rectangular
unless an unusual case is presented. by using a matrix size that is smaller in
2. Pulse sequence selection—The operator the phase-encoding direction than in the
usually has a choice of several pulse sequences frequency-encoding direction. This has the
which can be used to generate specific types additional benefit of reducing the total scan
of image contrast for various clinical situa- time by decreasing the number of phase-
tions. Spin echo and fast spin echo sequences encoding steps.
are typically used to obtain T1-weighted 6. Field of view (FOV)—Reducing the field of
and T2-weighted studies. MR angiography, view (FOV) increases the spatial resolution
cardiac imaging, and three-dimensional but may require more signal acquisitions to
studies most commonly use gradient echo maintain an adequate signal-to-noise ratio.
pulse sequences. Diffusion-weighted imaging, The FOV can also be selected to be rectagu-
contrast-enhanced perfusion studies, and lar by manipulating a parameter that is typi-
functional MRI employing blood oxygen cally called the “percent phase FOV.” For
level dependent (BOLD) contrast use echo- instance, if the FOV is selected to be 25 cm
planar imaging methods. and the % phase FOV = 80%, then the FOV
3. Timing selection—Three timing parameters will be 30 cm (FE) × 25 cm (PE). Together,
are of major importance. The repetition time the matrix size and the FOV determine the
(TR) is the time between RF excitation size of the pixels that consitute the image.
pulses. Reducing TR increases the T1 weight- 7. Number of signal acquisitions (NSA)—This
ing (T1W) of the image but also shortens the parameter is also called the number of
 CHAPTER 9  Magnetic Resonance Imaging Hardware 119

excitations (NEX) or number of averages defined through an interactive graphical user

(NA) on some systems. The more magnetic interface, in which the positioning and ori-
resonance (MR) signal acquisitions acquired, entation plan is presented as an overlay on
the better the contrast resolution because the localizer image data.
signal-to-noise ratio increases as the square
root of the NSA. Imaging time increases
Prescan Calibrations
directly as NSA increases.
8. Slice thickness—Reducing slice thickness Before each set of images is acquired, the system
improves the spatial resolution by reducing goes through an automatic prescan routine that
partial volume effects. The smaller the slice calibrates the pulse sequence to the patient’s
thickness, the longer the minimum possible body and the radio frequency coils being used.
TE will be. However, signal-to-noise ratio
also decreases proportionally with decreas- 1. Transmit/receive coil tuning—Patients are
ing slice thickness. electrically conductive masses that become
9. Receiver bandwidth (BW)—This important magnetically coupled when placed in the
parameter affects signal-to-noise ratio and RF coils used with MRI systems. The differ-
chemical shift artifact. The smaller the band- ences in the size and shape of patients
with, the more time there is between signal change the magnetic fields, altering the coil’s
samples. SNR is inversely proportional to resonant frequency and its electrical imped-
BW, but the minimum possible value of BW ance. Although, in principle, the resonant fre-
is often constrained by the desire to use short quency and impedance of the coil can be
values of TE. Some MRI systems do not tuned to match that of the patient for
allow direct selection of this parameter, but maximum RF signal reception, modern
all display its value. clinical RF coils typically are tuned in the
10. Magnetization preparation schemes— factory to accommodate a large range of
Presaturation pulses can be used to selec- patients. This strategy is employed to reduce
tively eliminate the signals from fat, or from setup time and maximize patient throughput.
water at specific locations. Magnetization However, the MRI system will perform an
transfer prepulses can be used to improve automatic check to determine that excessive
contrast in MR angiography images. Modern sensitivity has not been sacrificed for a given
MRI systems allow specification of a variety patient. Generation of a low-signal error
of presaturation schemes during parameter will likely neccesitate the operator changing
selection. to an RF coil that is more appropriate for
11. Reconstruction strategies—Modern MRI the body part being examined. If the error
systems also allow a variety of reconstruc- cannot be resolved an RF subsystem failure
tion options. Commonly available methods may be indicated.
include two-dimensional versus three- 2. Center frequency tuning—The MRI system
dimensional (3D) imaging, MR spectros- measures the resonant frequency at the center
copy, time-resolved imaging with contrast of the patient volume prescribed for imaging.
kinetics (TRICKS), and partially parallel Exact tuning establishes the slice position and
imaging. Usually advanced imaging pack- the resonance frequencies of both fat protons
ages must be installed on MRI systems for and water protons for the subsequent imaging
these types of options to be available to the process.
operator. 3. Pulse amplitude calibration—The flip angle
12. Slice positioning and orientation—The (α, 90°, 180°) is determined by the duration
number of slices, the gaps between the slices, of time that the RF is on and the amount of
and their relative orientations must all be power applied to generate the RF (pulse
120 PART III  The Imaging System

amplitude). Typically, a simple pulse sequence patient through a two-way audio system. The
that is sensitive to the RF pulse amplitude is operator warns the patient when the table is
used to determine the 90° RF pulse, and the about to move, instructs the patient about
other flip angles are then scaled to this value. when to take a breath hold and reminds the
4. Receiver gain calibration—This is an ampli- patient not to move during the study.
fier adjustment that scales the range of analog- 3. Image quality inspection—During the study,
to-digital conversion (ADC) values that are after an acquisition is completed, the operator
expected to be used to digitally encode should inspect the image data during the fol-
the received MRI signals. This calibration is lowing acquisition to ensure that image
used to ensure that the signal is not too low, quality and positioning of the images is ade-
resulting in noisy images, or too high, result- quate. If image quality is not adequate, then
ing in data clipping that produces image the acquisition can be repeated before the
distortion. patient is taken off the table.
5. Autoshimming—This process is invoked for
imaging methods that are sensitive to poor
Image Processing/
magnetic field homogeneity. A series of
Display/Manipulation
signal acquisitions occur, in which the slice-
selection gradient is applied, but no frequency- 1. Window width/level—This control is used to
encoding or phase-encoding gradient is used. set the contrast and shades of gray for the
This process generates a signal from the displayed image. The ranges are usually much
volume of interest. A simple algorithm is then wider for MRI than for CT.
employed to automatically adjust the current 2. Cursor on/off—This is used to place a cursor
that defines the X-, Y-, and Z-gradient zero on the image and provide for joystick, track-
values (gradient offsets) to maximize the ball, or mouse manipulation of the cursor.
signal. The pulsing following gradient offset 3. Region of interest (ROI)—ROI is used for
adjustment is repeated until the maximum calculation of area of measure and average
signal is achieved, at which time the autoshim and standard deviations of pixel values.
algorithm terminates and the imaging scan 4. Zoom—This is used to magnify the image.
proceeds. Some systems provide fixed magnification
factors, and others have continuously variable
factors. This is an electro-optical zoom.
Image Scanning Reconstructive zoom is not possible in MRI
1. Patient positioning—The patient is positioned as it is in CT.
on the table in the appropriate orientation 5. Profile/histogram—This plots the pixel values
(i.e., supine, prone, decubitis, etc.). Often along an identified axis as either a line graph
a surface coil or phased array coil is posi- or a histogram.
tioned and secured with respect to the 6. Highlight—This control selects pixel values
anatomy under investigation. The laser posi- within a given range for special attention.
tioning system is used to move the body part They may appear white or black or blink.
under investigation close to magnet isocenter. Special reconstruction algorithms are avail-
The patient is then moved to the position at able for surface rendering and volume
which the first, localizer images will be rendering.
acquired. 7. Display matrix format—This provides for
2. Patient monitoring—The operator must the simultaneous display of multiple images
monitor the patient at all times throughout or portions of an image on one video screen.
the data acquisition through windows and It is particularly helpful in MRI because so
TV. The operator communicates with the many images are acquired.
 CHAPTER 9  Magnetic Resonance Imaging Hardware 121

The computers interact with each other to

MAGNETIC RESONANCE
achieve the common objective of defining the
IMAGING COMPUTERS
imaging task, executing the pulse program, and
processing acquired MRI raw data. To provide
Types and Functions of MRI
simultaneous access to image data and increase
System Computers
image processing capabilities, a remote physi-
MRI systems typically use several different com- cian’s console, with its own computer, communi-
puters that perform a variety of functions impor- cates efficiently with the main console computer
tant to the MR imaging process. The computer and uses image data transfer protocols involving
at the scanner console has overall control of all the PACS for the display and image processing
processes but can pass specific tasks on to other functions.
specialized computers. The second type of computer used in MRI
For the computer to perform several tasks at systems is the acquistion control system. This
once—such as interpret operator input, pass computer is typically rack-mounted and located
instructions on for data acquisition, and perform with other systems electronics in the equipment
image display functions—the underlying control room. This system receives instructions from the
software, called the operating system, must be control computer (TR, slice thickness, number of
multitasking. A multitasking operating system slices, etc.) and converts them into specific pulses,
allows the computer hardware to be shared by timing delays, and amplifier gains. This com-
several programs and devices. puter is typically programmed in a language such
A good operating system minimizes the as C or C++, and comes equipped with compilers
impact of sharing by knowing when each and assembler routines that translate the pulse
program does not need to use the computer. An programs written in these higher-order languages
example occurs during the latter part of TR, into machine code, which is used to actually
when the acquisition program is just waiting for implement the pulse programs. This subsystem is
the nuclear spin system to recover to equilibrium. discussed in the following Sequencing System
Another example is the interval between when section.
the reconstruction requests the raw data of an It is also typical to have a third computer in
acquisition from the disk and when those data the MRI system that is dedicated to image recon-
are delivered. struction. This system must have high capacity
The MRI system operator interacts directly and exceedingly fast throughput. The image
with the control computer, which is often running reconstruction system is also usually physically
some version of the Microsoft Windows™ oper- present as a large integrated circuit board that is
ating system. The control computer is just a very mounted into one of the equipment room cabi-
powerful PC, with a lot of memory, storage nets. This system has a high capacity to store
space, a high-end graphics card, and fast busses data for manipulation because of the nature of
and processors. The control computer provides the MR signal and the number of signals required
an interface with the operator for control and for an image.
display functions, data archiving, and network The image processing computer must be fast
communications. Tasks within the MRI system to handle the high rate of data acquisition
are generally distributed between several sepa- and to accommodate the enormous number of
rate processing systems that are designed to opti- calculations required to produce an image. Image
mize execution of specific processes. processing times of less than 1 second are
required. These systems often employ one of
Distributed computation on an MRI system con- the many variations of the UNIX® operating
sists of several autonomous computers that com- system. Using dedicated processing and recon-
municate through a local computer network. struction computers, MR image processing is
122 PART III  The Imaging System

essentially simultaneous with most types of MRI image reconstruction method, computation of a
signal acquisitions. This is known as real-time single 256 × 256 image requires 512 × 256 fast
computing. Fourier transforms (FFTs). Each FFT requires
2048 complex multiplications, and these are the
most time-consuming parts of the FFT algorithm.
Storage Capacity A complex multiplication may be computed with
Although individual MR images are only of four real multiplications. Therefore the 256 ×
moderate size compared to CT images, clinical 256 2DFT image needs approximately 4.2 million
MRI protocols can produce quantities of data multiplications. In addition, advanced imaging
that are far in excess of those encountered with methods such as diffusion tensor imaging (DTI)
other medical imaging modalities. One three- and partially parallel image processing require
dimensional head image data set may yield 50 many more computation steps.
fast spin echo images at TEs of 120 ms for T2W For the maximum precision resolution in
images. the data to be preserved, the FFT should be
Each image can cover approximately 250 mm computed with floating-point numbers. A
FOV with a 512 × 384 matrix, each pixel 2 floating-point operation, such as a real multipli-
bytes, resulting in about 2.5 megabytes (MB) of cation, is referred to as a flop. The image data
data. A typical electrocardiogram-gated MR described above thus will require 4.2 million
cardiac examination may easily produce 1000 flops (MFLOPS). Reconstructing the image in 1
images of 256 × 256 pixels each. If each pixel second requires approximately 4 million flops
is 2 bytes deep, this amounts to over 16 MB per second (4 MFLOPS). Today images are
of data. reconstructed in fractions of a second, requiring
A typical personal computer in 2014 has a processing speeds rated in billions of flops per
disk storage capacity of perhaps 0.5 to 2 tera- second (GFLOPS).
bytes (TB) and a usable random access memory Only with the use of dedicated digital signal
(RAM) of 4 gigabytes (GB) to 8 GB. MRI com- processors (DSPs) and graphics processing units
puters tend to be at the high end of available (GPUs) can such a computational rate be
computers and fall into a class known as achieved. GPUs are dedicated electronic circuits
workstations. that are designed to manipulate and alter memory
Workstation consoles consist of a large, high- rapidly in order to accelerate the reconstitution
resolution display, a keyboard, and a mouse at a of images in a frame buffer intended for output
minimum. MRI system workstations typically to a display. Modern GPUs operate at speeds
provide support for error-correcting code (ECC) around 200 GFLOPS. Such speed is based on
memory, a larger number of memory sockets idealized situations; practical performance is
which use registered (buffered) modules, multiple slowed by limitations on the rate at which data
processor sockets, the most powerful CPUs may be moved on and off the disk and by data
that are commercially available, contain a high- bus speed.
performance graphics processing unit (GPU), High-performance MRI systems may incorpo-
and run a reliable operating system with advanced rate multiple DSPs and GPUs. The power used
features. by modern GPUs to efficiently manipulate com-
puter graphics, along with their highly parallel
structure, allows them to be more effective than
Computer Speed
general-purpose CPUs for algorithms that require
Not only must the storage capacity of the com- the processing of large blocks of data in parallel.
puter be large, but the computer must also In a workstation, the GPU can be present on a
perform computations quickly. In one common video card, on the motherboard, or even on the
two-dimensional Fourier transformation (2DFT) CPU. This technology, developed to support the
 CHAPTER 9  Magnetic Resonance Imaging Hardware 123

computer gaming industry, allows MR images to

Radio Frequency Amplifier
appear immediately after data collection.
The RF amplifer increases the signal output from
the frequency synthesizer so that adquate RF
MRI SYSTEM ELECTRONICS power can be applied to the RF coils. This must
The MRI system typically comes with several be done while maintaining amplitude/phase lin-
cabinets full of electronics that are housed in a earity and stability over dynamic ranges in excess
separate air-conditioned room next to the scanner of 60 dB (~1000 V).
room in which the gantry is located. Over the In order to implement slice selection efficiently,
years the size of these electronic components has pulse shaping circuitry also is employed to change
been shrinking so that what once required a the profile from a square RF pulse to a crafted
200 sq ft (~19 m2) room can now often fit in a RF pulse. The RF pulses produced for the MRI
space of 50 sq ft (~4.7 m2). Components are also system must not exhibit pulse droop for either
being made to be more compatible with strong short or long pulse conditions.
magnetic fields, and many of the electronic com- The system also should be immune to load
ponents are being designed to be more integrated changes when various coils and body parts
with the magnet gantry. No matter their size or are imaged. Because RF energy is directly pro-
physical location, the functions of these subsys- portional to RF frquency, MRI RF amplifiers
tems remain the same, and it is important to must have increasing power ratings when used
understand how they operate in the MR image with higher field magnets. Amplifiers used for
creation process. 1.5 T MRI systems have maximum outputs
ranging from 18 kW to 25 kW while those used
with 3 T MRI systems have outputs of 30 kW to
Frequency Synthesizer
40 kW.
The frequency synthesizer generates the funda-
mental resonance frequency used by the MRI
system to excite the magnetic nuclear spins. This
Gradient Coil Power Supply
device produces a low-level sinusoidal wave We have previously remarked on the fact that
signal covering a range of frequencies from a gradient coils create noise in a manner similar to
single fixed timebase or oscillator. Due to the loudspeakers. In early MRI systems the power to
extremely complex pulse crafting demands of the gradient coils came from digitally controlled
recently developed “RF shimming” or “parallel audio ampliers. In modern systems, power to the
excitation” schemes, modern MRI systems use gradient coils comes from specialized devices
direct digital synthesizers (DDS) that can create designed specifically for use in MR imaging.
arbitrary waveforms using frequency- and These gradient power supplies (GPS) operate
phase-tunable output signals, which are refer- in a current-driven mode that can accomodate
enced to a fixed-frequency precision clock source load inductances up to tens of Henries and load
(which is often referred to as the MRI system resistances in the range of 50 to 100 mΩ. Peak
clock). By using a procedure known as “divid- output currents can range up to 1 kA and peak
ing down,” the DDS architecture scales the fre- output voltages up to 500 V. In addition, the
quency clock using a programmable binary system has to be fast ramping and low noise with
tuning word that is typically 24 to 48 bits minimum waveform ripple. Often sold as a set,
long. This strategy enables a DDS implementa- a high-performance gradient power amplifier
tion to provide superior output frequency tuning system has three to six channels, for the X, Y,
resolution (on the order of microHertz) and and Z gradients and their associated shielding
subdegree phase tuning with minimum phase coil power supplies, which are preinstalled in a
noise. rack enclosure. For high-field MRI systems, an
124 PART III  The Imaging System

effective cooling system must also be installed to nano-ohm. Thus these magnets’ field strength
ensure temperature stability of the gradient coil will slowly drift down over time.
and power supply during scanner operation. Typical specifications for magnetic field drift
High-powered GPSs for whole-body systems are in the range of 0.01 ppm/hour. Thus the mag-
generally fall into one of three categories: netic field drifts down several thousand Hertz in
linear ampliers, switching systems, and resonant resonance frequency each year and periodically
systems. A linear GPS delivers power directly to the magnet power supply must be connected and
the coils, where most of the energy is subject to the B0 field “bumped up” so the magnet will be
ohmic losses. at a field appropriate for the operational fre-
Switching GPS systems and resonant GPS quency of the RF electronics.
systems use energy storage strategies to reduce
power dissipation. For the former, semiconduc-
Sequencing System
tor switches are employed to rapidly step between
current states; for the latter, the GPS is a compo- The sequencing system, also known as the pulse
nent in an L-C circuit that includes the gradient programmer, is a dedicated computer that is
coil as the inductor with the goal of producing usually found on the electronics rack in the MRI
sinusoidal current waveforms efficiently. In addi- system equipment room. This system is the traffic
tion, a pre-emphasis circuitry module, which is cop of the MRI scanner, taking instructions from
used to modify the gradient pulse waveform to the computer at the operator’s console and setting
minimize eddy currents, is often placed at the up the proper parameters for MRI signal acquisi-
input of the GPS. tion. The parameters are provided to the sequenc-
ing system, which plays out the correct MR pulse
sequence to acquire an image at the selected
Magnet Power Supply imaging plane.
The magnet power supply for a superconducting The sequencing system controls the RF trans-
magnet is used when the system is first installed. mitter and gradient amplifiers to cause an MR
This process is known as “ramping up to field.” image of the subject at an imaging plane to be
In order to accept current, the magnet coils must acquired. It also sets parameters such as band-
have some resistance. A superconducting heat width and amplification for the receiver subsys-
switch that is in parallel with the superconduct- tem. MRI pulse programmers used in commerical
ing coils—called a persistent switch—is engaged. MRI systems have specialized timing-generator
When a current is applied to the switch element, hardware units developed by the system manu-
it causes the switch to heat up, making it non- facturers. The instructions from the system
superconducting and allowing current to enter sequencer are often passed to the electronics in
the magnet coils. The power supply is pro- an assembly language, in which each statement
grammed to slowly inject current until the speci- corresponds to a single machine code instruction,
fied field has been attained. to enhance efficiency.
After the magnet is stabilized at the appropri-
ate B0 field, the current to the switch is turned
Digital Signal Acquisition System
off and it is allowed to cool back down to super-
conducting temperatures. The power supply can The NMR signal is detected by the receiving RF
then be removed, and the B0 field is maintained coils, often surface coils or phased-array coils,
by the current that continues to circulate in the which are discussed in Chapter 11. The detected
superconducting coils. Superconducting magnets signal then goes through a preamplifier, which is
present small amounts of resistance at the often located near the leads of the receiver coil.
leads. The total resistance in the coils of an atypi- The preamplifier increases the amplitude of the
cal 1.5 T magnet is less than a fraction of 1 MRI signal from tens to hundreds of mV to
 CHAPTER 9  Magnetic Resonance Imaging Hardware 125

several volts. The signal is then passed out of the ancillary devices that enhance certain functions.
magnet into the equipment room where it is digi- Power injectors, which are MRI compatible, are
tized by the analog-to-digital converters, pro- routinely used for examinations that use contrast
cessed, and filtered. agents, especially studies in which image acquisi-
With modern MRI systems using phased-array tion is coordinated with the injection timing.
coils, the digital signal system consists of multi- MRI-compatible infusion pumps are used to
ple RF channels, in each of which signal from a administer medications and other fluids during
discrete coil element goes through the same the MRI scan. Most MRI systems are also
process on separate hardware. For conventional equipped with an assortment of imaging test
MRI, typically two channels were used. With objects for system performance assessment. In
modern MRI systems employing partially paral- addition, pads and pillows are used to help posi-
lel imaging, up to 128 channels are being offered tion patients and on high-field MRI systems
commercially. special liquid gel pads mitigate magnetic suscep-
tibility artifacts. Finally, the cryogen filling equip-
ment, typically a set of nonmagnetic insulated
Digital Image Processing System
hoses and tubes with appropiate fittings, must be
A dedicated reconstruction computer is typically available for filling the magnet with liquid
employed to arrange the incoming digitized MRI helium.
signals and apply two-dimensional and three-
dimensional inverse Fourier transforms (FTs).
CHALLENGE QUESTIONS
Traditionally, the processing hardware has been
optimized for performing multidimensional FTs. 1. What are the three principal components of
However, sometimes data that are not acquired an MRI system?
with data points obtained on a rectilinear matrix 2. For any computer, especially those used
will require additional processing, such as inter- for MRI, what type of software program
polation to a grid. In addition, the partially par- controls the data acquisition and image
allel imaging that is standard on high-field MRI reconstruction?
systems requires several additional processing 3. What are the three types of MRI systems?
steps, which may include comparing signals from 4. In computer jargon, what is a flop?
two different coils, threshold masking, applying 5. The thumping noise generated in the
filters, extrapolation of the data, and applying gantry of an MRI system is due to what
complex fitting algorithms. Thus modern pro- component?
cessing engines employ a full range of digital 6. An MR image is reconstructed as a 512 ×
signal processing technologies including general- 512 matrix, 1 byte deep. How many mega-
purpose microprocessors, field-programmable bytes are required to store that image?
gate arrays (FPGAs), digital signal controllers, 7. What are cryogenic gases?
and stream processors, among others. 8. What are the primary functions of the MR
imaging process that often require separate
computers in the MRI system?
Ancillary Equipment 9. What is the main purpose of the iron yoke
for the MRI Suite in a permanent magnet imaging system?
In addition to the main subsystems of the scanner, 10. What does the word shim refer to for an
MRI systems are configured with a number of MRI system?
126 PART III  The Imaging System
 CHAPTER

10 
Primary Magnetic Resonance
Imaging Magnets

OBJECTIVES
At the completion of this chapter, the student • Discuss the terms cryogen, cryogenic gas, and
should be able to do the following: cryostat.
• Identify the properties of various types of • Convert temperature from one scale to
magnets used for magnetic resonance imaging another.
(MRI). • Identify differences between high-field and
• Describe superconductivity and its application low-field imaging.
to MRI.

OUTLINE
Permanent Magnets Superconducting Superconductivity
Electromagnets Electromagnets Superconductor Operation
Resistive Electromagnets

KEY TERMS
Cryogens Dewars Magnetic field homogeneity
Cryostat Fringe magnetic field Quench

The equipment necessary to perform magnetic The primary MRI magnet produces the
resonance imaging (MRI) is complex and sophis- static magnetic field (B0). Secondary magnets
ticated. A systematic walk through the major are used to shim the B0 field and produce gradi-
components of the MRI system helps to clarify ent magnetic fields (BX, BY, BZ). These fields
the system’s operation. are sometimes designated as GX, GY, and GZ.
A schematic block diagram of a typical MRI Finally, the radiofrequency (RF) system can
system is shown in Figure 10-1. The major com- also be considered as a secondary magnet. It
ponents of the gantry of an MRI system can be includes transmit and receive coils. The shim/
identified as either the primary magnets or the gradient coils and RF system are discussed in
secondary magnets. later chapters.

127
128 PART III  The Imaging System

Cryogen
storage

Magnetic power supply

Magnet Patient's table

Computer
cabinets

RF coils

RF electronics
Operating console Diagnostic
satellite console
FIGURE 10-1  Schematic diagram of the principal components of a magnetic resonance imaging system.

The primary magnet is the heart of the MRI

BOX 10-1 Types of Magnets Used
system. The function of the primary magnet is
to Generate B0
to provide a sustained, homogeneous B0 during
the MRI examination. The maintenance of a
homogeneous B0 is required because B0 homoge-
neity affects image resolution, uniformity, and
distortion.
At least two criteria in the selection of an MRI
system are driven by the type of primary magnet:
the desired field strength and siting limitations.
MRI, Magnetic resonance imaging.
Field strength and siting limitations are interre-
lated because for any given magnet design,
increasing field strength increases the size of the
associated fringe magnetic field. Fringe magnetic magnet, the main focus is on superconducting
fields are the component of B0 that extend outside electromagnets, the type used in more than 95%
the magnet and its housing (see Chapter 30). of installed MRI systems. A brief comparison of
Basically, two types of magnets are used to these systems is given in Table 10-1.
generate B0: permanent magnets and electromag-
nets. Electromagnets can be further classified as
PERMANENT MAGNETS
resistive electromagnets and superconducting
electromagnets (Box 10-1). Each magnet type has The permanent magnet is the most familiar type
advantages and disadvantages with respect to of magnet. This is the type of magnet used to
these two important, often overriding, criteria: demonstrate magnetic fields in science classes
field strength and siting limitations. Although and to fix paper to refrigerators. Permanent
the following discussion considers each type of magnets are components of compasses, motors,
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 129

TA B L E 1 0 - 1 Characteristic Features of Magnetic Resonance Imaging Systems

Permanent Resistive Superconducting

Magnet Electromagnet Electromagnet
B0 field intensity 0.1-0.3 T 0.2 T 0.5-4 T
Power requirements Low Very high Low
Cooling requirements None Water Cryogenic
Magnetic fringe field None Modest Strong

FIGURE 10-2  A permanent magnet magnetic resonance

imaging system featuring an open design provides easy
access and comfort for claustrophobic patients. (Courtesy
Hitachi Medical Corp.)

and audio speakers. They are inexpensive and

widely used for simple applications.

Permanent magnets occur naturally, or they can

be synthesized.

The earliest commercial magnets were made FIGURE 10-3  A method for rendering ceramic bricks
of iron and called ferrite magnets. In the 1930s magnetic with an electromagnet.
an alloy called alnico (aluminum, nickel, and
cobalt) was developed with a slightly higher
magnetic field than ferrite magnets. Alnico child during imaging. Open permanent magnet
magnets have been made with ever-increasing systems also make claustrophobic or anxious
magnetic field intensities. More recently, rare patients more comfortable. The magnetic field is
earth magnets have been introduced, which have typically produced by individual brick-size fer-
even higher magnetic field intensity. romagnetic ceramic materials that are rendered
Although magnetic field strengths of up to magnetic by charging them in the field of an
approximately 1.2 T can theoretically be achieved electromagnet (Figure 10-3).
with permanent magnets, field strengths of only Once magnetized, these bricks are then care-
about 0.3 T are practical for whole-body MRI fully oriented into an array, up to 1 m on a side,
systems due to their extreme weight. Figure 10-2 containing two to five layers. The fabrication of
shows a typical permanent magnet MRI system. such a large magnet made from smaller magnets
This design is often called an open MRI system is not a trivial task. The forces exerted are enor-
because it enables parents to remain with their mous, and if one brick is positioned incorrectly,
130 PART III  The Imaging System

contrary magnetic fields can result and cause the for head (50 cm) or whole-body (100 cm)
whole assembly to fragment violently. imaging. A typical permanent magnet design is
A variety of magnetic bricks are shown in shown in Figure 10-5. In such a design the four
Figure 10-4. Two assemblies are positioned corner posts are iron and provide a return path
opposite one another at a distance appropriate for the magnetic field.
A pole face is positioned on each magnet
assembly (Figure 10-6). These pole faces are care-
fully machined iron slabs designed to help orient
and shape the B0 field and to increase its homo-
geneity within the imaging volume. Often there
are adjusting screws or other mechanical shim-
ming devices to further refine the homogeneity of
the magnetic field after the imaging system is
installed in a prepared site.

Permanent magnets are shimmed with a pole

face.

An iron yoke is in physical contact with each

permanent magnet, and this contact serves three
purposes. First, it provides a mechanical frame
for assembly and stability. Second, it confines the
fringe magnetic field by concentrating the lines
of the magnetic field in this iron yoke. Finally, by
FIGURE 10-4  Individual magnets of varying sizes and
shapes are assembled to produce the strong magnetic field containing the fringe magnetic field, the yoke
of a permanent magnet magnetic resonance imaging also intensifies B0 in the imaging aperture.
system. (Courtesy Sumitomo Special Metals.) Without an iron yoke, B0 strength of such an

FIGURE 10-5  Typical permanent magnet design for imaging. (Courtesy Sumitomo Special Metals.)
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 131

Adjusting screw

Pole face

FIGURE 10-6  Permanent magnets are shimmed by positioning precisely machined pole faces.

B0

B0

A B
FIGURE 10-7  The presence of an iron yoke (A) with a permanent magnet imaging system intensifies the B0 field in
comparison to one without (B). The yoke provides a return path for the lines of the magnetic field.

assembly would be much less (Figure 10-7). The greater than 0.5 mT in any controlled area. This
yoke is usually made of soft iron laminated and level is chosen out of consideration for patients
bolted together like a transformer core. with cardiac pacemakers; it is not hazardous to
Table 10-2 presents the principal characteris- others.
tics of a permanent magnet MRI system. The The 0.5 mT field is within a few centimeters
principal advantage of a permanent magnet of the permanent magnet gantry because of the
MRI system is the insignificant fringe magnetic mass and design of the iron yoke. Other advan-
field, which for any MRI system must be no tages of a permanent magnet MRI system include
132 PART III  The Imaging System

TA B L E 1 0 - 2 Characteristics
Z
of a Permanent Magnet
Magnetic Resonance
Imaging System

Feature Value
Y
Magnetic field (B0) Up to 0.3 T
Magnetic field homogeneity 10-50 ppm
Weight 90,000 kg X
Cooling None
Power consumption 20 kW
Distance to 0.5 mT fringe field <1 m

low electric power consumption and the absence FIGURE 10-8  The long axis of the patient is the Y-axis
of a cooling system. in a permanent magnet magnetic resonance imaging
The principal disadvantage of a permanent system, corresponding to the axis identification of vector
diagrams.
magnet MRI system is the limited B0 intensity.
This places some restrictions on the type and
complexity of imaging allowed. Other disadvan-
tages include the relatively poor magnetic field requirement, and open architecture. However,
homogeneity, usually about 20 parts per million the use of permanent magnets in clinical MRI is
(ppm), and the excessive weight. Poor magnetic limited. This is because the maximum B0 inten-
field homogeneity results in reduced spatial and sity is low. Also the geometry of permanent
contrast resolution. magnets limits the efficiency of the gradient coils
The weight of a permanent magnet MRI used with them. Because of this, MRI systems
system limits its use to fixed sites. Although the based on permanent magnets compromise the
fringe magnetic fields tend to be small, the weight ability for some clinical applications, such as
of a well-designed permanent magnet MRI echo planar imaging. However, permanent
system can approach 90,000 kg (approximately magnet designs are being adopted for specialty
100 tons). This imposes significant mechanical MRI scanners, such as those dedicated to extrem-
considerations on the chosen location and usually ity imaging, because of their low cost and the
excludes all but ground floor siting. small fields of view used in these applications.
There is a fundamental difference between a
permanent magnet MRI system and one of elec-
ELECTROMAGNETS
tromagnet design. The B0 field of a permanent
magnet imaging system is vertical (Figure 10-8). Electromagnets make up the second type of
Therefore the Z-axis is vertical rather than hori- primary magnet system for imaging. It is useful
zontal as in most superconducting electromagnet to discuss resistive electromagnets and supercon-
imaging systems. The long axis of the patient in ducting electromagnets separately because they
a permanent magnet imaging system is the have significantly different operating characteris-
X-axis, and the lateral direction is the Y-axis. tics. However, the physics of image production
This corresponds to the axis identification in is exactly the same regardless of magnet type,
vector diagrams. especially when comparing equal B0. Neither the
Permanent magnets are attractive for low patient nor the secondary magnets can identify
magnetic field imaging applications because of the origin of the B0. There are no characteristic
their minimal fringe magnetic fields, low power image distinctions.
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 133

FIGURE 10-9  A resistive electromagnet usually has four separate coils to intensify B0 and make it uniform.

Resistive Electromagnets
B0
Resistive electromagnets are uncommon in
modern MRI systems and are typically found
only in older, legacy products. The B0 in a resis-
tive electromagnet imaging system is produced
by a large, classical electromagnet. All of the
early MRI investigations were conducted with
such resistive air-core electromagnets. The two FIGURE 10-10  This resistive electromagnet configuration
outside coils of the classical four-coil design are produces a vertical B0 field.
smaller in diameter, rendering the B0 more
uniform in the imaging volume between the two
large inside coils (Figure 10-9).
TABLE 10-3 Characteristics
One variation of a resistive electromagnet
of a Resistive
MRI incorporates the design shown in Figure Electromagnet Magnetic
10-10. This arrangement results in a vertical B0 Resonance Imaging System
field, which can be intensified by coupling to a
permanent magnet. Feature Value
The most common design of solenoid resistive
Magnetic field (B0) Up to 0.3 T
electromagnets uses aluminum strips wound spi- Magnetic field homogeneity 10-50 ppm
rally around a tube in several thousand layers. Weight 4000 kg
The advantage in cost and weight dictates alumi- Cooling Water, heat exchanger
num as the preferred resistive electromagnet Power consumption 80 kW
material. The mass density of aluminum is Distance to 0.5 mT fringe field 2 m
approximately one third that of copper; however,
it has only about 60% the conductivity of copper.
Table 10-3 gives the principal characteristics of Resistive electromagnet imaging systems
a resistive electromagnet MRI system. have some advantages. They are less expensive
to purchase than superconducting electromagnet
imaging systems because they operate at lower
The maximum field strength of the resistive elec-
field strengths and they do not require the preci-
tromagnet is approximately 0.3 T.
sion and homogeneity of a superconducting
134 PART III  The Imaging System

imaging system. Magnetic field homogeneity of high-energy physics now exceed 14 T. These sys-
10 to 50 ppm exists for most resistive MRI tems can achieve such high fields because of their
systems. Shimming this type of magnet is some- high electric current and small bore size.
what less difficult than shimming a supercon- High-field superconducting magnets have rel-
ducting system. ative advantages and disadvantages when com-
Because such a resistive electromagnet is pared with low-field systems. Higher B0 requires
readily brought up to designed magnetic field more intense gradient magnetic fields, resulting
strength, it is just as easily turned off. This in broader RF bandwidth.
removes the hazard of ferromagnetic projectiles Longitudinal relaxation time (T1) increases
during nonimaging time. It is a simple matter to with increasing B0 so that relatively longer repeti-
remove metallic objects that become attracted to tion times (TR) are required for the same T1
and stuck in the magnet. However, operation of weighting (T1W). B0 has no effect on transverse
resistive magnets is still expensive because they relaxation time, T2, at field strengths below
can consume large amounts of electricity. approximately 2 T. At higher B0, T2 decreases
Siting the resistive electromagnet imaging slightly.
system is easier than siting a superconducting The precessional frequencies of fat and water
electromagnet in terms of fringe magnetic fields. protons are separated by approximately 3.5 ppm.
A resistive electromagnetic imaging system also At 0.3 T this amounts to only 45 Hz and is not
weighs less than permanent magnet imaging noticeable. At 3 T the chemical shift is 450 Hz
systems (4000 kg versus 90,000 kg). For these and results in a distinct artifact. Artifacts caused
reasons, siting the resistive electromagnet imaging by magnetic susceptibility, blood flow, and
system is often the simplest. patient motion are more severe at high B0.
The principal disadvantage of this type of The most common magnets today use super-
imaging system is electric power consumption. A conducting technology to achieve intense, highly
resistive electromagnet imaging system is the homogenous magnetic fields. Table 10-4 presents
most power hungry of the three. A 0.2 T imaging the principal characteristics of a clinical super-
system may require 60 to 80 kW, and this is a conducting electromagnet MRI system.
continuous power drain when the magnet is on.
In addition, requirements for cooling the magnet
Superconductivity
must be met. Resistive electromagnets are water
cooled, with a closed primary loop communicat- At room temperature, all materials resist the flow
ing with a secondary single pass system through of electric current. Superconductivity is the
a heat exchanger.

Table 10-4 Characteristics

Superconducting Electromagnets of a Clinical
Electromagnets are created by conducting elec- Superconducting
tric current through coiled wire. MRI systems Electromagnet Magnetic
based on superconducting electromagnet tech- Resonance Imaging System
nology have the principal characteristic of high
Feature Value
B0 magnetic field strength.
Most superconducting clinical imaging sys- Magnetic field (B0) 0.5 T to 4 T
tems operate at 0.5 T, 1.0 T, or 1.5 T. Super­ Magnetic field homogeneity 0.1-5 ppm
conducting imaging systems at 3 T and 4 T Weight 10,000 kg
Cooling Cryogenic
are in clinical operation at many sites, but these
Power consumption 20 kW
are specialty systems. Superconducting electro- Distance to 0.5 mT fringe field 10 m
magnets used for analytical spectroscopy and
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 135

property of some materials that allows them to

BOX 10-2 The Four Electrical States
conduct electricity with no resistance. Such mate-
of Matter
rials reach superconductivity as temperature is
lowered to a critical temperature (T c). This criti- 1. Insulator
cal temperature varies with each superconduct- 2. Semiconductor
ing material. 3. Conductor
4. Superconductor
Superconductivity means that once the electric
current begins to flow, it will flow indefinitely.
NbTi

ELECTRICAL RESISTANCE
All clinical superconducting electromagnets
use niobium-titanium (NbTi) alloys that have a Cu
critical temperature of approximately 9 Kelvin TC
(K). Such a low temperature is achieved by
immersing the conductor in liquid helium. Liquid 0 10 20 30
K
helium vaporizes at 4 K and therefore easily
maintains the NbTi conductor below its critical Liquid He
temperature. Liquid nitrogen, which vaporizes at Room
77 K, was used to insulate the liquid helium. Liquid N2 temperature
Liquid nitrogen is seldom used in current cryo-
stats. Liquified gases that are used to keep the
conductors cold, such as liquid helium and liquid 0 TC 100 200
nitrogen, are called cryogens. TEMPERATURE (K)
Liquid He
The container housing the superconducting wire FIGURE 10-11  The electrical resistance of a conductor
and the cryogens is a cryostat. (Cu) and a superconductor (NbTi) as a function of tem-
perature. Tc, Critical temperature.

Superconductivity is one of the four electrical

states of matter (Box 10-2). In this state, elec- a new class of materials. They showed that the
trons flow in a conductor and experience no state of superconductivity could be made to exist
resistance. As the temperature of a conventional in some exotic materials at temperatures above
electrical conductor such as copper is lowered, 20 K. The materials showing the most promise
its resistance decreases (Figure 10-11). Theoreti- are compounds of lanthanum, barium, copper,
cally, the decrease is linear to the origin, so its and oxygen.
value is 0 at 0 K. Since then, other investigators have demon-
A superconductor such as NbTi has higher strated the state of superconductivity at even
resistance at room temperature than copper and higher temperatures (Figure 10-12). However,
is therefore a poor conductor at room tempera- these materials are ceramics with compositions
ture. Its resistance decreases linearly with tem- more akin to chalk than to ductile metals such
perature just as copper does. However, when it as copper. Additionally, these materials lose
reaches its T c, its resistance immediately drops their superconducting properties rather easily
to zero. when placed in a high magnetic field. Neverthe-
There is much scientific investigation of super- less, the science research on high critical tempera-
conductivity at this time that may profoundly ture superconductors continues because they
affect the future of MRI. In 1987 Bednorz and would be of great benefit to the electrical trans-
Müller won the Nobel Prize in physics for their mission industry and reduce the cost of MRI
work on high-temperature superconductivity in systems.
136 PART III  The Imaging System

320
300 Room temperature
280
Water freezes
260
?
Temperature (K)

160 HgBaCaCuO
Freon
140
ThBaCaCuO
120
BaSrCaCuO
100
YBaCuO
80 Nitrogen
60
40 NbTi NbGe LaBaCuO
Pb Nb NbN NbSn
20 Hydrogen
Helium Hg
0
1900 1920 1940 1960 1980 2000
Year
FIGURE 10-12  Recent years have shown a dramatic rise in the critical temperature for superconducting materials.

There are several advantages to superconduct- strength of 1 T, the 0.5 mT (5 gauss) fringe mag-
ing MRI magnets. The high magnetic field inten- netic field associated with pacemaker exclusion
sity is essential if spectroscopy is planned; extends some 10 m in all directions.
however, the tolerances on field homogeneity are The extent of the fringe magnetic field may be
more restrictive at such high B0 intensities. reduced by passive or active shielding. Passive
Higher B0 intensity is desirable because the shielding consists of placing ferromagnetic mate-
increased magnetic resonance (MR) signal from rial in the walls of the examination room or
such imaging systems produces higher signal-to- around the magnet (Figure 10-13). However, this
noise ratio (SNR), allowing shorter examination passive shielding is heavy (as much as 250 tons
times. In addition, because of the increased SNR, for a 7 tesla magnet) and expensive. Therefore,
higher spatial and contrast resolution images can passive shielding is rarely used for clinical MRI
be obtained with these systems. systems.
The field of the superconducting electro­ The active shielding approach uses a second
magnet can also be homogenized or shimmed to set of superconducting coils positioned outside
a degree not achievable with the other magnet the primary coils but still inside the cryostat with
systems. The solenoidal design of superconduct- electric current moving in the opposite direction
ing magnets makes them inherently homoge- of the primary current. This produces a magnetic
neous. Shimming is important for small field that counteracts the primary magnetic field,
field-of-view (FOV) imaging, optimal fat sup- resulting in a reduced fringe magnetic field
pression, fast imaging, and spectroscopy. outside the magnet.
Homogeneity of less than 1 ppm over a 40 cm Proper design of active shielding can also
FOV is obtainable with superconductive shim- improve the magnetic field homogeneity, and
ming. At 1 ppm, a 1 T imaging system has a B0 this feature has allowed superconducting magnets
of 1 T ± 1 µT (see Chapter 11). to be shortened from lengths of 2 m down to
The principal disadvantage of these systems is arround 1.5 m. Thus, although active shielding
their intense fringe magnetic field, which can is expensive and may appreciably change the
compromise site selection. At a magnetic field weight of the system, it is generally standard
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 137

FIGURE 10-13  A passively shielded superconducting imaging system. (Courtesy Siemens Medical Systems.)

now. Figure 10-14 shows a cross-sectional image-guided interventions such as biopsies are
diagram of a representative superconducting required. If actively shielded, the 0.5mT fringe
magnet. The 20 K reflective shield and vacuum magnetic field is restricted to 4 m.
thermally uncouple the container of liquid helium As mentioned above, one disadvantage of the
(lHe) from the environment. Typically, six short, superconducting magnet is the need for cryo-
main magnet superconducting solenoid coils genic gases. Maintenance of the NbTi supercon-
would be immersed in the lHe and once powered ducting wire in a superconducting state requires
up they can sustain current indefinitely as long an environment below its critical temperature
as they are maintained at a low temperature. of 9 K. The NbTi wire is wound on aluminum
The two active-shielded coils are designed to formers that are immersed in liquid helium,
reduce the extent of the fringe magnetic fields which vaporizes at 4 K (Figure 10-15). External
into the scanner room. The differences in the forces, principally in the form of thermal radia-
number of lines on each coil suggest how the tion, will cause the liquid helium to heat up and
design varies the current densities in the coil vaporize over time. Because liquid helium is
windings, which is important to the design of expensive, its vaporization is reduced by sur-
very short magnets. The service turret at the top rounding the helium compartment with concen-
is where the superconducting leads are attached tric insulating compartments (Table 10-5).
to power the magnet up and where the thermally In older magnets there were two intermediate
insulated tubing is inserted for filling the magnet layers of super-insulation that had to be kept at
with cryogens. 20 K and 80 K, respectively. The compartment
A more recent family of superconducting containing the coils and liquid helium is sepa-
electromagnets has been developed to compete rated from an outer vessel containing liquid
with the “open architecture” of permanent nitrogen by a vacuum shield (see Figure 10-16,
magnet imaging systems. Such imaging systems A). Liquid nitrogen, which vaporizes at 77 K,
are designed around two magnets stacked so that is used in these magnets instead of liquid
the B0 field may be 1.0 T with an imaging aper- hydrogen (20 K) or liquid neon (27 K) because
ture of 50 cm. This geometry can be useful if it is less expensive. Furthermore, because of its
138 PART III  The Imaging System

Service turret Vacuum

container

Active
LHE shielding
coils

Active shielding coils

Six magnet coils

20° K
shield
LHE Dewar

FIGURE 10-14  This cross-sectional diagram of a typical superconducting magnet shows a standard coil configuration.
All of the coils are located inside a dewar, or thermally isolated vessel, containing liquid helium (lHe).

high boiling point, liquid nitrogen also vaporizes elimination of He boil-off through recirculation
more slowly. also allows reduction of dewar size and complex-
ity through elimination of one thermal shield,
Temperature Scales enabling a more compact and open design for the
magnet. This arrangement of multiple thermal
TC = 5/9 (TF - 32) compartments is shown in Figure 10-16, B.
TF = 9/5 TC + 32 However, as magnet sizes decrease and the
TK = TC + 273 spatial resolution of MRI improves, the vibra-
where the subscripts C, F, and K refer to Celsius, tions caused by the oscillating displacer of the
Fahrenheit, and Kelvin, respectively. cryocooler may be a problem with some MRI
applications.
Modern magnet design eliminates the nitrogen Figure 10-17 shows an actively shielded 1.5 T
compartment in favor of superior vacuum com- large-bore MR imaging system with a cryocooler.
partments with lHe replenishing devices called These several thermal shields are designed to
cryocoolers. In modern magnets, cryocooler maintain the superconducting condition of the
systems are routinely used to cool the radiation electromagnet. They shield the main magnet coil
shields of the main magnet system and to recon- windings from the gradient fields so that eddy
dense the helium present in the main magnet currents (see Chapter 11) will be induced in
system. Cryocooling alone can reduce lHe con- them, rather than disrupt the function of the
sumption from 1 l/hr to less than 0.2 l/hr. The main magnet coils. Generally a cryocooler, the
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 139

Superconductor Operation
Once the superconducting electromagnet is
energized or “brought up to field,” it no longer
requires external electrical power to maintain
the B0. This dramatically reduces the power
requirements of the site relative to resistive
 electromagnets.
However, the disadvantage of the supercon-
Liquid He ducting system is that the coil windings must be
(4.2 K) maintained at cryogenic temperature, which
requires the occasional replenishing of liquid
helium. This replenishing must be performed by
trained service personnel.

Advances in magnet design have reduced helium

boil-off and eliminated liquid nitrogen.
~10° Temperature How then is the MRI system connected to a
FIGURE 10-15  This diagram shows how the resistance power source that is resistive in nature and the
of the superconducting magnet coils (Ω) increases as a magnet brought up to designed magnetic field
function of temperature. When the critical temperature of
liquid helium (4.2 K) is exceeded, the finite resistance in
strength in the superconductive state? A direct
the coils causes large voltages (kV) to be generated. current (DC) is supplied to the magnet through
a resistive power supply. The primary coil has a
shunt across it, but there is a heater on the shunt,
causing it to behave as a resistive conductor
TA B L E 1 0 - 5 Notable Temperatures
(Figure 10-19). Electron flow is from the resistive
and Characteristic Effects
power supply and through the primary coil. As
Temperature (K) Characteristic Effect this electric current exits the primary coil to com-
plete the circuit, it senses the high resistance of
0 Absolute zero; all motion stops
4 Helium vaporizes
the shunt.
9 NbTi superconducts When the electric current designed to produce
77 Hydrogen vaporizes a given B0 is attained, the shunt heater is turned
273 Ice melts off and the shunt becomes superconducting.
293 Room temperature is achieved Now the electric current must choose between
373 Water boils the resistive power supply and the superconduct-
ing shunt. The shunt wins, and the power supply
is unplugged.
essential component of the unit known as a cold If the superconductor becomes resistive, it
head, recondenses helium vapors back into the heats up. As it heats, it becomes more resistive.
liquid helium bath. Typically the cold head is The result of such a situation is a quench.
placed on top of the magnet, next to the input Quenches may occur as a result of a lack of
ports for liquid helium filling, the receptacle for liquid helium or mechanical trauma to the cryo-
the demountable leads, through which current stat that causes the conductor to come in contact
is applied to the magnet coils (Figure 10-18). with a warmer component of the cryostat. If this
Advanced cryostats of actively shielded magnets contact happens, an uncontrolled quench occurs,
can go up to 3 years without helium refill. and the energy stored in the magnet is converted
 Liquid nitrogen

Liquid helium

Cold head Cryocooler

80K

4.2K 4.2K

Magnet with liquid Magnet with IHe

A He and liquid N B and cryocooler
FIGURE 10-16  Older magnets (A) used two layers of cryogens; liquid helium (lHe) and liquid nitrogen (lN). With the
development of cryocooler technology, a modern magnet (B) contains only lHe. This allows the overall size of the magnet
to be reduced and/or the magnet bore to be made larger.

FIGURE 10-17  A 1.5 T actively shielded superconducting magnet designed with an integral cryogenerator. (Courtesy
Philips Medical Systems.)
 CHAPTER 10  Primary Magnetic Resonance Imaging Magnets 141

to heat. This causes additional warming of the Much of this heat causes the cryogens to boil
coil, taking more sections over their critical tem- off. However, some of this heat may raise the
perature, which in turn will release more heat—a temperature of the cryostat and melt the super-
positive feedback system. The quench results in conductive windings. Such a violent quench can
the B0 magnetic field collapsing and the liquid even destroy a magnet. Even if the magnet sur-
helium rapidly boiling off. vives the quench, re-energizing the magnet can
be an expensive process.

A quench occurs when a superconducting magnet

Helium vent to warms and the electromagnetic coil becomes
Helium exterior resistive.
recirculation
system
In order to prevent an excessive release of
energy a “controlled quench” can be initiated
that “ramps down” the magnet and shunts the
electrical energy to a dummy load to shut down
the individual coils in a coordinated manner
that will avoid damaging the magnet (Figure
10-20). When a quench occurs, pressure builds
up within the dewar and “bursting discs” are
often employed that blow out under high pres-
sure and channel the large volume of cold helium
gas out of the magnet and often into the atmo-
FIGURE 10-18  A helium recirculation system, including
the cryocooler, is shown here on top of the magnet. The sphere outside the building. (Numerous videos of
venting system gives the helium gas a clear path to the magnets quenching are available for viewing on
exterior in the event of a quench. YouTube.) Oxygen level monitors are sometimes

Heater
coil

Super
conducting
Superconducting
switch
coil

Liquid
helium
Helium
dewar

FIGURE 10-19  This schematic depicts the circuitry in the service turret, which allows the magnet to receive current. In
order for current to flow initially, the circuit must have some resistance. This is provided by the heater coil, which causes
a small section of wire to be resistive and take on current. Once the current has increased to the point that the desired
magnetic field strength has been established, the superconducting switch is used to short-circuit the magnet. Then the
current circulating in the superconducting coils is isolated and the power supply can be removed from the magnet.
142 PART III  The Imaging System

1 2 3 4 5 6 Cryogenic gases are supplied to the MRI facil-

ity in large, thermos-like vessels called dewars.
The dewars require special handling and storage.
Only specially trained service personnel should
replenish cryogens because of the potential
hazards to skin, which can be burned severely by
contact with a cryogen.
Current (A)

Coils 3 and 4

CHALLENGE QUESTIONS
Coils 1 and 6 1. What is the difference between an electro-
magnet and a permanent magnet?
Coils 2 and 5
2. What happens when a superconducting
magnet quenches?
3. What level of magnetic field intensity (B0)
can be obtained with a permanent magnet,
Time (s) resistive electromagnet, and superconduct-
FIGURE 10-20  A controlled quench allows the six ing electromagnet?
magnet coils (upper right insert) to “ramp down.” Coils 4. State the critical temperatures for NbTi
heat up at different rates when the quench control cir-
cuitry is engaged to ensure that the heat produced by superconducting wire, liquid helium, liquid
energy transferring from the magnetic field is distributed nitrogen, and water.
evenly throughout magnet. The helium gas is then vented 5. A comfortable room temperature is consid-
to the exterior. ered to be 70° F. What is that temperature
expressed in degrees Celsius and Kelvin?
installed in the magnet room to raise an alert 6. Superconducting MRI magnets are made
if an excessive amount of He leaks from the more homogeneous with electromagnetic
exhaust pipe. shim coils. How is this accomplished with a
There are safety concerns in the event of an permanent magnet imaging system?
unplanned quench. One may first hear a hissing 7. What is so special about absolute zero?
noise caused by the release of helium, nitrogen, 8. List some advantages to the use of a resistive
or both. Everyone should be immediately evacu- electromagnet for MRI.
ated because both gases can displace oxygen and 9. Describe the four electrical states of
cause asphyxiation. Then, as you are dying, your matter.
voice will sound like that of Donald Duck. 10. Define superconductivity.
 CHAPTER

11 
Secondary Magnetic
Resonance Imaging Magnets

OBJECTIVES
At the completion of this chapter, the student • Discuss the difference between homogeneous
should be able to do the following: and inhomogeneous radiofrequency (RF)
• Identify shim coils and their purpose. probes.
• Understand the shape and positioning of the • List the advantages and disadvantages of
three sets of gradient coils. surface coil and phased array coil imaging.
• Describe quadrature detection and its
relevance to magnetic resonance (MR) signal
detection.

OUTLINE
Shim Coils Y Gradient Coils (GY, Gϕ) Surface Coils
The Parts per Million Scale Combined Gradients Phased Array and Matrix
Shimming the Magnet The Radiofrequency Probe Coils
Gradient Coils Quadrature Coils
Z Gradient Coils (GZ, GSS) Body Coils
X Gradient Coils (GX, GR) Head/Extremity Coils

KEY TERMS
Decoupled Gradient magnetic field Slew rate
Eddy current Shimming Slice selection gradient

On the one hand, magnetic resonance imaging gradient magnetic fields are produced by second-
(MRI) requires a uniform static magnetic field ary electromagnetic coils that are usually of the
(B0) to produce a strong, coherent free induction resistive type but can be superconducting.
decay (FID) signal. On the other hand, precisely In addition to the gradient coils that produce
fashioned gradient magnetic fields are required the gradient magnetic fields, there are shim coils
to deliberately spoil the homogeneity in order in some high-field MRI systems. There also is
to provide spatial location information. The an integrated radiofrequency (RF) coil, which is

143
144 PART III  The Imaging System

often referred to as “the body coil.” These sec- homogeneity of ±1 ppm is a variation of ±1 µT
ondary coils are positioned inside the bore of the throughout the imaging volume of a 1.0 T magnet
cylinder of the gantry. (1 T = 106 µT). A homogeneity of ±5 ppm in a
Indeed, the gradient coils and shim coils are 1.5 T magnet would be ±7.5 µT.
designed to produce precise magnetic fields,
which are superimposed on the primary static
Magnetic field homogeneity is specified as the
magnetic field. The body coil most typically is
ppm over a specific spherical volume.
used to transmit RF energy into the patient, but
sometimes may be used to receive the induced
MRI signal from the patient too. Acceptable homogeneity over a 20 cm spheri-
cal volume is approximately 20 ppm with perma-
nent magnets but only 1 ppm with superconducting
SHIM COILS magnets. The spherical volume is usually speci-
Shimming is the process of making the B0 field fied in terms of the diameter of the spherical
uniform throughout the imaging volume. The volume or DSV.
shimmed volume will determine the volume of For a 1.5 T field, the proton Larmor frequency
the static magnetic field suitable for imaging. The is 63 MHz. If field homogeneity of 10 ppm is
shimming process often involves placing steel achieved, the resonance frequency in the shimmed
bars in the magnet at precise positions in trays volume will be 63 MHz ± 630 Hz. For a 0.5 T
located along the inside of the magnet bore. system operating the same way, 10 ppm homo-
Primary shimming of a magnet is usually per- geneity means the proton Larmor frequency is
fomed at the factory. The MRI system installa- 21 MHz ± 210 Hz.
tion engineer will likely do some fine tuning of
the magnet’s shim at the site.
Shimming the Magnet
Room temperature coils may also be used to
improve the B0 field unformity within the patient The imaging volume in an MRI system generally
for particular procedures that require high mag- consists of a cylinder with a diameter of approxi-
netic field homogeneity. On systems operating at mately 60 cm and a length along the Z-axis of
1.5 T and below, this procedure only requires a 50 to 70 cm. A conventional six-coil supercon-
DC offset of the current applied to the gradient ducting magnet will normally produce a mag-
coil. At 3 T and higher, separate coils are installed netic field with homogeneity of approximately
that adjust the higher functional orders of the ±5 ppm.
magnet field. These are the shim coils. When determining the B0 magnetic field homo-
geneity, the installation engineer will typically
use a small probe containing a tiny sample of
Magnetic field uniformity is termed homogeneity.
water to measure the resonance frequency and
map the magnetic field. A special jig is used to
position the small probe at precise locations
within the magnet bore.
The Parts per Million Scale Once the measurements are performed, then
The specification of magnetic field homogeneity the homogeneity can be modified by placing steel
as parts per million (ppm) is borrowed from nu- shims in trays located in the bore of the magnet.
clear magnetic resonance (NMR) spectroscopy. The magnetic field can be mapped again and
This allows field homogeneity to be expressed this procedure is repeated until satisfactory mag-
independent of field strength. If a 1.0 T magnet netic field homogeneity is achieved. Figure 11-1
were perfectly homogeneous throughout the im- is a diagram that shows the geometry of this
aging volume, it would be stated as ±0 ppm. A measurement.
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 145

Off-axis

On-axis
Isocenter
Away from
isocenter

FIGURE 11-1  This diagram shows the positions (blue lines) on the prescribed spherical diameter where the installation
engineer makes measurements of the magnetic field homogeneity using a small radiofrequency probe that measures the
local resonant frequency.

Once the magnet has been properly shimmed, accomplished with the aid of superconducting
the procedure is not repeated unless operating coils within the cryostat itself.
techniques or the site environment changes Room temperature or resistive shim coils are
greatly. With high-speed functional imaging tech- configured to alter the magnetic field on the basis
niques and chemical shift–sensitive fat suppres- of spherical coordinates. The “first order” terms
sion, better B0 homogeneity is required. are the familiar x, y, and z directions. Second
Current superconducting magnets are designed order terms include x2, y2, z2, xy, yz, and so on.
and manufactured according to precise specifica- These “higher order” terms are required only for
tions. Gross shimming is performed at the factory extremely high-homogeneity applications, such
with the installation engineer completing only as spectroscopy.
minor corrections. What used to take up to The most common method for application-
2 weeks now requires just hours. specific shimming in conventional MRI systems
is to offset the baseline currents of the X, Y, and
Shimming is the process of making the B0 field Z gradient coils. This method can usually achieve
homogeneous. the desired degree of field homogeneity with the
standard hardware.
An actively shielded magnet incorporates Shimming using only the gradient coils
superconducting shim coils in the cryostat with produces adequate results partly due to the supe-
the primary magnet. This seems to be the most rior homogeneity of modern superconducting
effective shim design of all. Field homogeneity of magnets. However, for higher field systems that
less than ±1 ppm is attainable with supercon- are used for spectroscopy studies, in which a
ducting shim coils. higher degree of homogeneity is required, an
During installation of an MRI system the independent set of room temperature shim coils
magnet is brought up to field, and the homogene- may be necessary.
ity of the B0 field is optimized with the bore During prescan calibration, additional shim-
clear of gradients or RF coils. For superconduct- ming may be required because of patient size,
ing electromagnets, this homogeneity may be shape, and magnetic susceptibility. The patient
146 PART III  The Imaging System

Z Coils

Y Coils

X Coils
FIGURE 11-2  The positioning of the three sets of gradient coils.

may reduce the field homogeneity due to the (up to 40 mT/m). Gradient magnetic fields are
differences in magnetic susceptibility among dif- measured in millitesla per meter, and when fast
ferent tissues. For SE imaging, such prescan imaging is required, bigger is better.
shimming often is not required. However, at high These coils must be able to switch on and
B0 with fast imaging and especially for magnetic off rapidly. Switching times of less than 500 µs
resonance (MR) spectroscopy, prescan shimming are necessary for many MRI applications.
may be necessary. Such prescan shimming is Advanced gradient coils, such as those used for
computer controlled and requires little technolo- echo planar imaging, can produce gradient fields
gist time. of up to 40 mT/m, with switching times of as
little as 50 µs.
GRADIENT COILS
The maximum rate of rise to maximum gradient
If room temperature shim coils are used, another
amplitude is the slew rate. The time required to
cylinder is positioned inside the cylinder on
switch gradient coils from 10% to 90% of the
which the shim coils are wound. The gradient
maximum rated gradient strength is known as
coils are positioned on this second cylinder. If
the rise time.
room temperature shims are not used, the gradi-
ent coils are just inside the bore of the primary
magnet, near the walls of the housing. A typical slew rate curve is shown in Figure
There are three sets of gradient coils, one each 11-4. This representation shows that the
for the X, Y, and Z directions. Figure 11-2 shows maximum gradient field, 40 mT/m, is obtained
the configuration of such coils. Gradient coils are in a rise time of 250 µs. Therefore the slew rate
not coils of wire in the normal sense. Instead, is 160 T/m/s.
they are broad, thick, copper conducting bands
(Figure 11-3). Slew Rate
These bands are referred to as conductors and 40 mT/m in 250 ms
typically measure 10 mm wide and 4 mm thick. 40 mT/m/250 ms
The large size is necessary to reduce resistance
4 ¥ 10 -2 T/m/2.5 ¥ 10 -3 s
and carry the intense electric current (up to 30 A) 160 T/m/s
required to generate the gradient magnetic fields
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 147

FIGURE 11-3  Gradient coils consist of large, band-type conductors.

40mT/m
40 Slew rate =
160T/m/s

30
Gradient
amplitude Slew rate = 30T/m/s
(mT/m)
20
15mT/m
Rise time
= 250µs
10
Rise time = 500µs

0 100 200 300 400 500

time (µs)
FIGURE 11-4  The slew rate is the time required to energize or switch off a gradient coil.

Because the current in the gradient coils exceptionally important because it is the limiting
cannot be brought to the required amplitude specification for short repetition time (TR), short
instantaneously, the waveforms used are gener- TE, and total imaging time.
ally trapezoidal, as shown in Figure 11-5. The switching of the gradient coils produces
For routine SE imaging, slew rate is important the “thump, thump, thump” in the imaging aper-
because it limits the minimum time-to-echo (TE) ture that is heard by the patient. When a gradient
that is available. For fast imaging, slew rate is pulse begins, the electric charges accelerate in the
148 PART III  The Imaging System

Unidirectional gradient pulse

Bipolar gradient pulse

FIGURE 11-5  The representation of gradient pulses in a pulse sequence diagram.

gradient coil. Accelerating electric charges in a coils from moving when they are energized and
magnetic field will experience a force that is at dissipates the heat generated by the gradients.
right angles to both the direction of current flow The casing also can be designed to muffle the
and the magnetic field. When the gradient is sound of the gradient switching for patient
switched off, the conductor will experience a comfort.
force in the opposite direction. The alternate
pushing and pulling on the conductors of the The Lorentz force, resulting from accelerating
gradient coils produces noises in the same way charges in the gradient coils, causes the thump-
that loudspeakers make sounds. ing sound heard in MRI systems during scannning.
At low gradient pulsing rates a thumping
sound is produced, but at the higher gradient MRI systems incorporating high slew rate
pulsing rates used for MRI a nearly pure tone gradient coils induce eddy currents more effi-
can be produced. These sounds are of adequate ciently. The rapidly changing gradient magnetic
quality that a few demented researchers have field can induce eddy currents in anything nearby
been inspired to develop musical arrangements (patient, coils, probes), but most important, in
that can be played on MRI scanners. the conductive structures of the interior of
Because the gradient coils are experiencing the annulus of the cryostat, which can lead to
the Lorentz forces, they also tend to torque or image artifacts and a loss of spatial and contrast
twist. If this happens without constraints, the resolution.
coils change shape, and the resulting magnetic The principal method for controlling eddy
field is altered. Therefore an important design currents is the use of shielded gradient coils. The
criterion for the gradient coil conductors is that gradient shielding coils are located near the
they should be firmly attached to the former magnet bore walls and are designed to produce
tube so that the resulting motion is on a micro- eddy currents that counteract the eddy currents
scopic level and the vibrations produced by the produced by the main gradient coils. Thus the
whole tube structure do not result in physical gradient magnetic fields are effectively shielded
or mechanical degradation of the whole MRI from producing eddy currents at the surface of
system. the cryostat.
For the prevention of physical distortion, the Self-shielding gradient coils also can be
gradient coils often are imbedded in a strong designed to improve the linearity of the overall
epoxy resin casing. This casing prevents the gradient field in the patient. The deployment of
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 149

X Coils

Z Coils
FIGURE 11-6  Z gradient coils change the gradient mag-
netic field along the Z-axis. This field, BSS, is often used to
select a transverse section of the patient for imaging.
FIGURE 11-7  X gradient coils produce a gradient mag-
netic field across the patient. For transverse imaging, this
field is frequency encoded and is often called the readout
gradient magnetic field (BR).
self-shielded gradient coils allows the design of
gradient coils that produce large imaging volumes
even in the shorter superconducting magnets.
Because gradient coils are designed to have The Z gradient magnetic field is symbolized by Bz
low resistance, heating does not really become a or Bss, for slice selection.
problem until echo-planar imaging (EPI) speeds
are approached. Sometimes water-cooling is The BZ coils are much more efficient than the
used. BX and BY coils because all segments of the coils
contribute to the useable gradient magnetic field.
For the BZ coils there is a larger gradient per
Z Gradient Coils (GZ, GSS) ampere-turn.
The Z gradient coils are usually a pair of circular
coils, each of which is wound on the cylinder at
X Gradient Coils (GX, GR)
opposite ends of the imaging volume (Figure
11-6). These are often called Maxwell coils, As shown in Chapter 2, the magnetic field lines
named after James Clerk Maxwell. If a direct of a circular coil of wire are along the axis of
current with opposite polarity is passed through that coil. As a result, the X and Y gradient coils
the two coils, a small change in the magnetic field are more difficult to fabricate and position on the
along the Z-axis of the gantry is produced. The cylinder because they cannot be made as circles.
currents used for SE imaging are approximately Figure 11-7 shows the way the X gradient mag-
20 A, producing a linear change in main mag- netic field is induced by a pair of coils—actually
netic field strength of 10 to 40 mT/m. four saddle-shaped coils in sets of two—
This change in magnet field strength allows positioned on either side of the cylinder.
for selection of a slice along the axis of the gantry. By convention, these coils are positioned so
When transverse slices are required, the Z gradi- that the gradient magnetic field is across the
ent is the slice selection gradient. The stronger patient laterally. The axis is therefore the hori-
the Z gradient electric current is, the stronger zontal axis across the patient from side to side.
will be the Z gradient magnetization, and this These coils behave in precisely the same way
will result in thinner slices. as the BZ coils. Direct current of opposite polarity
150 PART III  The Imaging System

Y Coils

FIGURE 11-9  When all three gradients are energized at

FIGURE 11-8  Y gradient coils produce a gradient mag- the same time, an oblique plane is imaged.
netic field through the patient from front to back. For
transverse imaging, this field is usually the phase-encoding
gradient magnetic field, Bϕ.
Together, the Y and X gradient allows precise
determination of where the contribution to the
is applied to produce a gradient magnetic field. MR signal from each voxel originated within the
The X gradient current and induced gradient transverse imaging section.
magnetic field are similar in magnitude to the Z
gradients. For transverse images, the Y gradient magnetic
The X gradient magnetic fields provide spatial field is usually phase encoded and symbolized as
localization along the X-axis, side-to-side across BY or Bϕ.
the patient, and can also be used for slice selec-
tion (sagittal images), phase encoding, or fre- The previous discussion is an instructional
quency encoding. convenience that correctly applies to images of
the chest and abdomen. Transverse images of the
For transverse images, the X gradient magnetic head are usually acquired with the phase-encoded
field is usually frequency encoded and symbol- gradient running from left to right. With this
ized as Bx or BR, for the read gradient. technique, the number of phase-encoding steps
can be reduced to produce a rectangular field of
view that better conforms to the elliptical shape
Y Gradient Coils (GY, Gϕ) of the head’s cross section.
A gradient magnetic field along the Y-axis of
the patient is produced by a set of coils that look
Combined Gradients
and operate exactly like the X gradient coils
(Figure 11-8). By convention, the Y-axis is the Magnetic fields add vectorially. When energized
vertical axis through the patient in the anterior- simultaneously, the currents in the three pairs
posterior direction. of gradient coils do not produce three separate
The Y gradient is normally, though not neces- magnetic fields but rather a single composite
sarily, used for phase encoding the MR signal magnetic field.
during transverse imaging and is symbolized as All three gradients are energized simultane-
BY or Bϕ. It can also serve to perform slice selec- ously to obtain an oblique image (Figure 11-9).
tion (coronal images) and to frequency encode. If the current through each pair of gradient coils
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 151

RF coil

XY plane

FIGURE 11-10  A simple, circular radiofrequency (RF) coil can be used with a vertical field permanent magnet imaging
system because the coil axis is in the XY plane.

is precisely controlled, the plane for imaging can manufacturer considers its specific RF technol-
be precisely specified. ogy to be an important trade secret.
The gradient coils are under coordinated elec- The signal generated for the RF probe comes
tronic control during the MRI examination by from a device called a frequency synthesizer. This
the pulse programmer to achieve the desired gra- is the master frequency source for the MRI
dient magnetic fields. Oblique sections can also system. It provides a tunable frequency band
be imaged with three-dimensional Fourier trans- from which the Larmor frequency can be accu-
formation (3DFT) techniques. rately determined for each individual examina-
tion. This small but precise signal is then
magnified by an amplifier that feeds the RF
THE RADIOFREQUENCY PROBE energy into the coil.
The RF probe is essentially a coil of wire not The simplest RF probe is a coil of wire
unlike the gradient, shim, and primary coils of wrapped around a patient or placed on the body
the MRI system. It differs, however, in that but separated by a covering. The intensity of the
it must accommodate a high-frequency alternat- emitted RF signal and the sensitivity to the signal
ing current of 10 to 200 MHz, depending on B0 received from the patient are maximum in a
intensity, so that it can produce a radio signal at volume approximately equal to the diameter of
the Larmor frequency. the coil.
Furthermore, some RF probes must be pre- Outside of this coil diameter, signal intensity
cisely designed to behave as both a transmitter and sensitivity decrease rapidly. Such a simple
and receiver of RF. The design of the RF probe type of coil is easily adaptable to a permanent
is one of the more critical engineering features magnet imaging system (Figure 11-10). This
of an MRI system. Such design has developed arrangement adds to the simplicity of a perma-
into a very exact science and each MRI system nent magnet imaging system and improves
152 PART III  The Imaging System

signal-to-noise ratio (SNR) at a given magnetic Initially, the most widely used design for the
field strength. RF probe was a saddle coil (Figure 11-12, A).
For most superconducting electromagnetic In such a coil, the intensity of emitted RF and
imaging systems, the primary static magnetic the sensitivity of the received signal are nearly
field is along the axis of the patient. The trans- uniform within the confines of the coil. The
mitting and receiving RF coils must produce degree of such uniformity results in great measure
magnetic fields that are perpendicular to the from the precise spacing of the loops of the
main magnetic field. To be used, the simple cir- saddle. However, this type of coil has low signal
cular coil has to penetrate through the patient sensitivity.
either laterally or anteroposteriorly (Figure
11-11). Because patients object to such treat-
Quadrature Coils
ment, other coil designs and shapes had to be
devised. Quadrature coil design improves SNR by detect-
ing the MR signal from multiple directions, as
shown in Figure 11-13. The coils view the signal
as though they were a pair of stereo lenses. The
result is better sensitivity to the MR signal rela-
tive to linear coils.
Quadrature coils are also constructed to be
more homogeneous for RF transmission and
reception. For this reason, quadrature coils have
replaced the saddle design for virtually all homo-
geneous applications (e.g., head, body, knee).
There are several types of quadrature coils, which
are all much more complicated than the saddle
coil. One version, the “birdcage” resonator, is
widely used (see Figure 11-12, B).
There are basically two types of RF probes.
Homogeneous volume coils are typically used
both to transmit RF and to receive the MR signal.
FIGURE 11-11  Radiofrequency probes for electromag-
netic imaging systems are complicated because a simple These include the head and body coils and other
circular coil would have to pass through the patient for its special application coils, such as an upper
axis to be in the XY plane. extremity/shoulder coil shown in Figure 11-14.

A B
FIGURE 11-12  A, The saddle coil. B, The “birdcage” resonator type of quadrature coil.
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 153

Mxy Mxy

Mxy Mxy

FIGURE 11-13  Quadrature detection improves signal-to-noise ratio (SNR) with multiple pairs of coils.

Head, body, or extremity coils are known as technologist may excite and subsequently detect
homogeneous coils; surface coils are referred an MR signal from a slice through the liver, but
to as inhomogeneous coils. the noise detected by the body coil comes from
the liver, chest, and abdomen within the body
coil. This reduces the SNR because of the higher
Most inhomogeneous coils are surface level of noise.
coils, and the word inhomogeneous is used All coils except the body coil improve SNR by
because these coils do not transmit RF in a reducing the detected noise. Tissue not being
homogeneous fashion. For this reason, they examined is eliminated from the coil’s sensitive
usually receive only. When these coils are used, volume. For example, the head coil does not
the RF is transmitted by the head or body coil. detect noise from the liver because the liver is far
The transmitting coil is then electrically silenced from the coil (i.e., outside its sensitive volume).
or decoupled while the surface coil receives the The other coils also improve SNR through
signal. increased signal detection by having the signal
The MR signal detected from a patient con- better “fill” the coil volume.
sists of two components: signal and random
noise. The signal comes from only the slice of
Body Coils
tissue being excited.
The noise detected by the RF coil comes from The most robust RF probe is the body coil. This
all the tissue within the sensitive volume of the statement is justified because the body coil can
coil. For example, with the body coil, the MRI image any part of the human anatomy. The body
154 PART III  The Imaging System

A B

C D
FIGURE 11-14  A basic set of radiofrequency coils would include coils to image the head, body, extremities, and spine.
These coils are for: A, spine and soft tissue neck studies; B, chest, abdomen, and pelvis imaging; C, imaging across the
glenohumeral joint; D, soft tissue, skull base, neck, and brain imaging. (Courtesy GE Medical Systems.)

coil is wound on a former and is just inside the thoracic/lumbar surface spine coil to produce
gradient coils. The body coil is always a transmit/ images of the total spine.
receive device (Figure 11-15). However, it is
often possible to design coils for imaging specific
Surface Coils
anatomy with better SNR and thereby produce
better images than those produced with a single For most imaging systems, the RF probe serves
body coil alone. as both the transmitter of the RF and the receiver
of the MR signal. One notable exception is the
surface coil.
Head/Extremity Coils A surface coil is a specially designed coil that
A common alternative to the body coil for is usually flat but can also be other shapes and
imaging cranial anatomy is the head coil. An is used to obtain high SNR images of anatomy
extremity coil is available to obtain high- close to the surface. The disadvantage of its use
resolution images from the lower extremity. A is reduced field of view (FOV). The surface coil
typical quadrature birdcage head coil (Figure may be encased in a rubberized or plastic matrix
11-16) can be attached to a neck coil and a to make it somewhat pliable or in a hardened
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 155

FIGURE 11-15  A body probe of a quadrature detection phased array design to image the cervical, thoracic, and lumbar
spine simultaneously. (Courtesy Hitachi Medical Systems America.)

FIGURE 11-16  Quadrature design, birdcage, and head coil shown with oncology mask. (Courtesy Midwest RF.)
156 PART III  The Imaging System

A B C
FIGURE 11-17  Typical surface coils from small circular (A) to rectangular coils for spine (B) to large rectangular
(C) available for magnetic resonance imaging. (Courtesy Siemens Medical Systems.)

composite material for increased stability. The

coil is placed on the surface of the patient at the
anatomical region under investigation.
Surface coils come in many different sizes
and shapes and are usually fabricated for specific
anatomy. A minimal complement of surface
coils consists of several flat, circular devices of
varying diameters and a “license plate” coil for
spine imaging.
Other surface coils include individual probes
designed specifically for extremities, joints, or
orbits (Figure 11-17). Special coil designs now
include breast (Figure 11-18), prostate (Figure
11-19), and almost any other organ of interest.
A high-resolution prostate image is shown in
Figure 11-20.
When in use, the surface coil is positioned
inside the head or body RF probe on the patient
but is insulated from the skin. These devices are FIGURE 11-18  Bilateral breast coil for magnetic
resonance mammography. (Courtesy Siemens Medical
normally used only as MR signal receivers and
Systems.)
rely on the head or body coil to transmit the RF.
This type of use requires that the transmitter and
surface coils be decoupled from one another so Surface coils provide better contrast resolution
that they do not interfere with sensing the weak and better spatial resolution.
MR signal or that one does not “intercept” the
RF transmission of the other. A surface coil has improved contrast resolu-
Lack of decoupling can cause the receiving tion because of higher SNR. The higher SNR
coil to heat up, possibly destroy the surface comes from two processes. The signal is increased
coil electronics, and potentially burn the patient. because the coil is placed closer to the body part
Special hardware is often used to actively decou- being imaged. The noise is reduced because noise
ple the two coils by detuning one of the coils is not received from the portions of the patient
while the other is operating. that are not intended to be in the image.
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 157

coil is smaller than the head or body probe, it

has a smaller sensitive volume, which results in
a restricted FOV. This is acceptable if only a
small region of anatomy is to be imaged.
However, positioning the surface coil requires
more time. To maximize the signal from the
small volume, the MRI technologist must pay
closer attention to patient setup. The surface
coil must be as orthogonal or perpendicular to
the XY plane as possible for maximum sensitiv-
ity. Just a slight tilt in the coil can result in sig-
nificant loss of signal. With experience, MRI
technologists learn to position surface coils easily
and quickly.

Surface coils have limited FOV.

Phased Array and Matrix Coils

FIGURE 11-19  Endorectal coils for prostate imaging. The principal objection to the use of surface coils
(Courtesy Siemens Medical Systems.) is the limited FOV. This deficiency has been over-
come with the development of phased array or
Since the SNR is higher, protocols with multicoil systems. If several surface coils are con-
increased spatial resolution can be used with nected and positioned so that they have minimal
adequate overall image quality. A surface coil coupling, a single large image can be produced.
image of the cervical spine, for instance, has a Each coil must have its own receiver channel for
pixel size of 0.4 × 0.4 mm for a 10 cm FOV and processing that single image of the cumulative
a 256 matrix (Figure 11-21). Because a smaller sensitive volumes of all the coils (Figure 11-22).
volume of tissue is being imaged, the pixel size In this situation, the SNR and spatial resolu-
for a given matrix with a surface coil is always tion of each individual surface coil is maintained
less than that for a whole-body RF probe. in the composite image. For example, if four
surface coils are used to image the thoracic and
Spatial Resolution/Pixel Size lumbar spine, each having a 10 cm FOV acquired
at a matrix size of 256 × 256, the composite
FOV = 10 cm = 100 mm image could have up to a 40 cm FOV with a 256
Matrix size = 256 ¥ 256 × 1024 matrix (pixels are added only along the
100 mm coils). Current routine applications of this tech-
Pixel size = = 0.39 mm
256 nology include imaging of the spine (Figure
2 pixels per line = 0.78 mm/lp 11-23), pelvis, and breast.
Hence: spatial resolution = (0.78 mm/lp)-1 Phased array coils offer several advantages.
= 1.28 lp/mm First, when the signals from the various coil ele-
ments are combined through averaging, a much
= 12.8 lp/cm
more uniform image can be produced over a
large imaging volume. Second, either linear
Disadvantages of surface coil imaging include receiver or quadrature receiver principles can be
limited FOV and positioning. Because the surface used with phased array coils, which can result in
158 PART III  The Imaging System

FIGURE 11-20  High-resolution, T2-weighted, three-dimensional fast spin echo (FSE) image of the prostate acquired
with an endorectal coil and pelvic phased-array showing carcinoma (low intensity) on the left. (Courtesy William Bradley,
San Diego, CA.)
 CHAPTER 11  Secondary Magnetic Resonance Imaging Magnets 159

increased SNR. Phased array coils also enable Large phased arrays called matrix coils have
parallel imaging methods in which the MR been implemented on some commercial MRI
imaging process can be made more quickly scanners. These matrix coils can contain 32, 64,
instead of gaining greater SNR. Parallel imaging or more individual coils. Unlike phased array
is dicussed in detail in Chapter 22. coils, which are usually designed for specific
body parts, matrix coils have very general appli-
cations. Electronic switching circuitry in the
matrix coils is interfaced to the MRI scanner
electronics, which allows the operator to select
which coil elements to use for a particalar study.
Matrix coils are often used with parallel MRI
methods.

CHALLENGE QUESTIONS
1. What is the purpose of shim coils in a super-
conducting MRI system?
2. Surface coil imaging results in better SNR
and better spatial resolution. What are its
principal limitations?
3. A superconducting MRI system operating at
3 T is said to have a field homogeneity of
±1 ppm. How should that be expressed in
millitesla?
FIGURE 11-21  A high-resolution image of the cervical 4. A surface coil with an 8 cm FOV is used
spine. (Courtesy Larry Rothenberg, New York, NY.) to image the orbit with a 512 × 512

COIL COIL COIL COIL

Digital Receiver 1
Combined
Digital Receiver 2 reconstruction

Digital Receiver 3

Digital Receiver 4 COMPOSITE

IMAGE
FIGURE 11-22  An array of four surface coils positioned to image the entire spine.
160 PART III  The Imaging System

reconstruction matrix. What is the limiting

spatial resolution for this examination?
5. A gradient magnetic field has a maximum
amplitude of 25 mT/m, with a rise time of
100 ms. What is the slew rate?
6. What can be done to improve the spatial
resolution of an MRI system?
7. For sagittal plane imaging, which gradient
coil must be energized (GX, GY, GZ)?
8. Surface coils are said to result in better con-
trast resolution. Why?
9. For a transverse image, is the read gradient
magnetic field (BR) frequency encoded or
phase encoded?
10. What is a homogeneous RF coil?

FIGURE 11-23  Composite image of the entire thoracic

and lumbar spine acquired with multiple surface coils.
(Courtesy GE Medical Systems.)
PART
IV
Image Formation
 CHAPTER

12 
Digital Imaging

OBJECTIVES
At the completion of this chapter, the student • Distinguish between bits, bytes, and words.
should be able to do the following: • Discuss the requirements on signal sampling
• Discuss the difference between binary and to avoid the aliasing artifact.
decimal number systems. • Identify spatial frequency domain (k-space).
• Convert binary numbers to decimal and • Describe spatial localization in a magnetic
decimal numbers to binary. resonance (MR) image.

OUTLINE
The Computer’s View of the Spatial Localization and Three-Dimensional Fourier
World Magnetic Resonance Transform Magnetic
Binary Number System Imaging Resonance Imaging
Bits, Bytes, and Words Projection Reconstruction
Quantization or Resolution Magnetic Resonance
Sampling Imaging
The Spatial Frequency Domain Two-Dimensional Fourier
Image Matrix Transform Magnetic
Frequency Domain Map Resonance Imaging

KEY TERMS
Byte Dynamic range Sampling
Digit False contouring
Digital imaging Pixel

When it is applied to medical imaging, the term x-rays do with a radiographic intensifying screen
digital imaging implies that a digital computer is phosphor. The MR signal does not present infor-
used and that the image is composed of discrete mation in a form that can be directly viewed
picture elements, that is, pixels. electronically, as does ultrasound.
Magnetic resonance imaging (MRI) could The MR signal provides information about
not be performed without the digital computer the spatial frequency content of the image rather
because the magnetic resonance (MR) signal does than directly about the spatial positioning of
not interact directly with a viewable medium as information in the image. Figure 12-1, A, is a

162
 CHAPTER 12  Digital Imaging 163

A B

C D
FIGURE 12-1  A, A plane image of two balls. B, The sinogram of the two balls. C, The magnitude of the spatial frequency
representation of the image. D, The phase of the spatial frequency representation of the image. C and D together are
the data set that is actually received in magnetic resonance imaging.

computer-generated image of two bright balls on raw data from such an MR image is much more
a dark background. Figure 12-1, B, is a set of difficult than understanding that from a projec-
projections called a sinogram from that image of tion radiograph or a CT sinogram.
the two balls. This type of data set is obtained
from a parallel beam x-ray computed tomogra- The MR image is acquired in the spatial frequency
phy (CT) image. The positions and the objects domain (k-space).
may be inferred at least to a limited extent from
the projection data because the data are in the The digital computer converts this informa-
spatial domain, which means that the data are tion about the spatial frequency domain into the
related directly to coordinates in space. spatial domain of the MR signal within the
The magnitude of the spatial frequency of the patient to produce the image. In this chapter,
image is shown in Figure 12-1, C. Figure 12-1, some characteristics of the digital computer
D, shows the phase data from the representation are reviewed and the details of the formation of
of the two balls. Because the MRI data are in the MR images are examined to explain why MRI is
spatial frequency domain, making sense of the inherently digital.
164 PART IV  Image Formation

TABLE 12-1 Organization of Binary

Number System

Decimal Binary Binary

Number Equivalent Number
0 0 0
1 20 1
2 21 + 0 10
3 21 + 20 11
4 22 + 0 + 0 100
5 22 + 0 + 20 101
6 22 + 21 + 0 110
7 22 + 21 + 20 111
8 23 + 0 + 0 + 0 1000
9 23 + 0 + 0 + 20 1001
10 23 + 0 + 21 + 0 1010
11 23 + 0 + 21 + 20 1011
12 23 + 22 + 0 + 0 1100
13 23 + 22 + 0 + 20 1101
FIGURE 12-2  The origin of the decimal number system 14 23 + 22 + 21 + 0 1110
is obvious. 15 23 + 22 + 21 + 20 1111
16 24 + 0 + 0 + 0 + 0 10000

THE COMPUTER’S VIEW OF

THE WORLD
converting alphabetic characters, decimal values,
The digital computer works with numbers, spe- and logical functions to binary values.
cifically, binary numbers. The binary number set In the binary system, 0 is 0 and 1 is 1. The 1
consists of 0 and 1. Perhaps if humans had only is actually 20. Any number raised to the zero
2 fingers instead of 10, the binary number system power is 1; therefore 20 equals 1. In binary nota-
would seem to be more natural than the decimal tion, the number 2 equals 1 times 21 plus no 20.
number system. This is expressed as 10.
A single number in the decimal system is The decimal 3 equals 1 times 21 plus 1 times
called a digit, as are the fingers shown in Figure 0
2 or 11 in binary form. The decimal 4 is 1 times
12-2. A single number in the binary system is 22 plus no 21 plus no 20 or 100 in binary form.
called a bit, which is short for binary digit. As shown in Table 12-1, each time 2 is raised to
The digital computer uses the binary number an additional power to express a number, the
system because it is so easy to implement with number of binary digits increases by 1.
real-world components. A 1 is simply defined to Easily recognizing the powers of 2 is essential.
be any voltage above a specified value and 0 to Power-of-2 notation is used in radiologic imaging
be any voltage below a specified value. to describe image size, image dynamic range
(shades of gray), and image storage capacity.
Table 12-2 reviews these power notations. In
Binary Number System both power notations, the number of zeros to the
In the binary number system, counting is done right of 1 equals the value of the exponent.
by counting 0 and 1 and then over again (Table
12-1). There are only two binary digits, 0 and 1, Question: How is the number 193 expressed in
and the computer performs all operations by binary form?
 CHAPTER 12  Digital Imaging 165

possible number of characters. Depending on

TA B L E 1 2 - 2 Power of 10, Power of 2,
the computer, a string of 8, 16, 32, 64, or pos-
and Binary Notation
sibly some other number of bits is manipulated
Power of 10 Power of 2 Binary Notation simultaneously.
10 = 1
0
2 =1
0
1
To encode is to translate from ordinary characters
101 = 10 21 = 2 10
102 = 100 22 = 4 100 to computer-compatible characters (binary
103 = 1000 23 = 8 1000 codes).
104 = 10,000 24 = 16 10000
105 = 100,000 25 = 32 100000 Bits are grouped into bunches of 8 called
106 = 1,000,000 26 = 64 1000000 bytes (B). Computer capacity is expressed by the
27 = 128 10000000 number of bytes the memory of the computer
28 = 256 100000000 can accommodate. For example, 1 kilobyte (kB)
29 = 512 1000000000 is 210 or 1024 bytes; 1 megabyte (MB) is 220 or
210 = 1024 10000000000 1,048,576 bytes.
Prefixes such as kilo, mega, and giga are not
metric in computer use but refer to the nearest
power of 2. Popular personal computers have
Answer: Because 193 falls between 27 and 28, it 32-bit or 64-bit microprocessors and contain
will be expressed as 1 followed by 7 binary 2 to 4 GB of main memory. The workstations
digits. Simply add the decimal equivalents of and minicomputers used in radiology also have
each binary digit from left to right. main memory capacities of gigabytes, and these
Yes 27 =  1 =  128 values are continuously increasing. Secondary
Yes 26 =  1 =  64 memory in the form of magnetic and optical
No 25 =  0 =  No 32 disks can have memory measured in terabytes
No 24 =  0 =  No 16 (TB) (1 TB = 240 = 1,099,511,627,776 bytes).
No 23 =  0 =  No 8
No 22 =  0 =  No 4 Question: How many bits can be stored on a
No 21 =  0 =  No 2 4 MB chip with a byte size of 8 bits?
Yes 20 =  1 =  1 Answer: 1,048,576 bytes/MB × 8 bits/byte ×
11000001  =  193 4 MB
or
Digital medical images are made of discrete 220 bytes/MB × 23 bits/byte × 22 MB
numerical values arranged in a matrix. The size 33,554,432 bits
of the image is described in the binary system of
numbers by power-of-2 equivalents. The most Depending on the computer configuration, 1,
popular MR image matrix sizes are 256 × 256 2, or 4 bytes usually constitute a word. In the
(28 × 28) and 512 × 512 (29 × 29). case of a 16-bit microprocessor, a word is 16
consecutive bits of information that are inter-
preted and shuffled about the computer as a unit.
Bits, Bytes, and Words Each word of data in memory has its own
The computer will use as many bits (zeros and address. Often each byte is addressable.
ones) as necessary to express a decimal number. The number of bits determines the resolution
The 26 uppercase and 26 lowercase characters or precision of the numbers that the computer
of the alphabet and other special characters manipulates. This means that the number of bits
are usually encoded by 7 bits. The use of 8 bits determines the total number of elements that the
by newer extended character sets doubles the computer can count.
166 PART IV  Image Formation

This is similar to the U.S. Postal Service Zip A voltage range of 24 V is typically used in
Codes. When the Postal Service wanted to add MRI. Thus, 12 bits (212 = 4096) are required
resolution to the Zip Codes, it added four more because 12 bits allow a range from 0 to 41.96 V;
numbers at the end of the five-numeral Zip Code 11 bits (211 = 2048) provide a range of only 0 to
to define the carrier route and the postal zone to 20.48 V, which would not be sufficient.
which a letter is addressed. A total of 16 or 24 bits are commonly used to
An 8-bit computer number can count from 0 allow extra bits for computations. For example,
through 255 and thus can represent 256 (28) dif- a 22 V signal and 23 V signal are added; 12 bits
ferent values. A 16-bit number can count from 0 would not be enough to hold the result, 45 V. An
through 65,535 and thus can represent 65,536 additional bit is required. Although more than
(216) different values. ample, computers with word lengths of 32 and
Regardless of the number of bits the computer 64 bits are becoming more popular as the prices
uses, it is limited in the number of distinct values of the computers decrease because they typically
that it can represent. It cannot hold continuous offer other attractive features such as faster
data the way a tape recording, a photograph, or computation.
a radiograph can. In the digital computer, limited Dynamic Range.  One area in which the resolu-
resolution and limited storage affect the manipu- tion of the data affects medical images is in the
lation of data. MRI is limited in the same way. number of gray levels used to display an image.
This is referred to as the dynamic range or
gray scale resolution of an imaging system. An
Quantization or Resolution imaging system that displays only black or white
Precision.  The limited number of elements the has a gray scale resolution of 21 (1 bit) or 2. Such
computer can count restricts the precision with an image is of very high contrast but displays
which the computer can store values. For proper little information. Although the value of each
interpretation, some quantities must be specified pixel is unimportant, the number of values is
with great precision. However, imprecision is a extremely important in determining the final
fact of life, and people live comfortably with image.
acceptable degrees of imprecision.
When someone asks, “How far is it to Dallas?” The range of pixel values in an image is the
a satisfactory answer is to the nearest mile or so. dynamic range or gray scale resolution.
An answer to the nearest foot or inch is unneces-
sary. When a person gets a driver’s license, the In MR images, gray scale resolution refers
person’s height is required to the nearest inch, to how small a difference in pixel intensity cor-
not the nearest quarter inch. Age is not given responds to a 1-bit difference in pixel value.
precisely to the minute but usually only to the Visually, gray scale resolution refers to the
year. Such limits of precision are totally satisfac- number of different shades of gray that can be
tory in everyday life. Considerably more preci- perceived.
sion is required for MRI. The gray scale resolution of the human eye
In the design of a computer system, the is somewhere between 24 (4 bits) and 25
required resolution of the data is determined (5 bits) or 16 to 32, shades of gray, stretching
first. For example, differences of 0.01 V may be from black to white. The gray scale resolution
distinguished in an MR signal—be it a free induc- of the MR signal from the patient exceeds 210
tion decay (FID), a spin echo (SE), or a gradient (10 bits).
echo (GRE). An 8-bit computer word allows for Although humans cannot perceive such a fine
256 discrete steps or a voltage range from 0 to gray scale resolution, a computer with sufficient
2.55 V. A 16-bit computer word gives a voltage capacity can store that much information. The
range from 0 to 655.35 V. finer the steps within the span from black to
 CHAPTER 12  Digital Imaging 167

white, the more gradual the gray scale represent- 1000

ing the range from minimum to maximum inten-
sity. Therefore the finer the gray scale resolution, 120
the better the image. 500
Early image display devices were capable
of only 4 bits (24) or 16 gray levels. Current
120 Window
displays are generally not limited to shades of 20 Window level
0 width
gray but provide “true color” in 24 bits: 8 bits
to represent red, 8 bits to represent blue, and
8 bits to represent green. Each of these colors
has 28 = 256 levels which can be combined to −500
produce a total of 16,777,216 mixed colors 20
(256 × 256 × 256).
In CT, the Hounsfield units have a range of
−1000
2000. In radioisotope scintigraphic imaging,
thousands of counts in a pixel may occur in a FIGURE 12-3  Postprocessing an image for various com-
binations of window level and window width allows the
first-pass study, and in MRI, T1 and T2 values entire gray scale to be viewed.
may have ranges of several thousand millisec-
onds. Despite this, the human visual system can
resolve no more than 32 shades of gray, and so
in a display system, somewhere between 32 and That loss may be perceived as false contour-
256 gray levels are sufficient to ensure no loss of ing, in which boundaries appear in the image
visual information. where there should be none. However, gray scale
However, just as CT images are displayed with resolution is not the only limitation implicit in
bone or soft tissue windows to evade the mis- working with digital computers. Limited storage
match between the resolution in Hounsfield units capacity also leads to limitations on spatial
and the gray scale resolution of the observer, the resolution.
32 or so perceptible shades of gray in the human
visual system must be used to full advantage. Question: How many bytes are required for a
This is done by adjusting the brightness and con- 256 × 256 MR image having a 12-bit dynamic
trast of the MR image so that all the gray scale range?
information available can be perceived. Answer: 12 bit ÷ 8 bit/byte = 1.5 byte/pixel
The full dynamic range of a 12-bit display is 256 × 256 × 1.5 = 98,304 bytes
visible with window level and window width or 98 kB
postprocessing (Figure 12-3). With variation of
the window level and width, the entire 4096 gray
scale can be viewed for diagnosis.
Sampling
It is necessary to consider an MR image dis- When the computer stores data, it must be able
played with different numbers of gray levels to to address the data. In the same way that there
understand the significance of the concept of an are a limited number of postal zones in the United
acceptable degree of imprecision. In Figure 12-4, States that can have a unique five-digit Zip Code,
the same image is displayed with 8, 7, 6, 5, 4, 3, there are a limited number of memory locations
2, and 1 bit of gray scale resolution. These bit that can be used for image storage. This limita-
values correspond to 256, 128, 64, 32, 16, 8, 4, tion is imposed by the limited number of bits in
and 2 shades of gray, respectively. With an insuf- the memory address. Consequently, for the image
ficient number of gray levels available, there is a to be stored in the computer, the continuous
visual loss of gray scale resolution. object to be imaged must be sampled.
168 PART IV  Image Formation

A
256 128

64 32

C
FIGURE 12-5  A, A continuous free induction decay (FID).
B, A sampled version of the FID. The computer stores only
the values taken at the regularly spaced indicated points
along the horizontal axis. C, The connect-the-dots recon-
struction of the original signal from the sampled values.
16 8

instance, the sampled data set is similar to a

connect-the-dots picture.

The sampling process limits the rate at which a

continuous signal is digitized.

Sampling is a familiar process. The business

4 2 pages of the newspapers give closing prices for
stocks each day. Even though the price of a stock
FIGURE 12-4  Different gray scale resolutions in an image may change during the day, only its price at the
display. For many people, the loss of gray scale resolution
end of the day and the range of its price during
does not become visible until only about 16 shades of gray
are present. At that point, apparent edges or contours are the day are given.
seen in the image where there should be none. A television picture is not a continuously
changing image, but rather, a rapid succession
of still images displayed at the rate of 30
The process of sampling involves taking occa- images each second. Obviously, there is no loss
sional values, or samples, of information from of important information about the action
the MR signal. This is illustrated in Figure 12-5, on the television screen, even though only 30
in which a continuous FID is represented by a set samples a second are seen, rather than continu-
of values taken at a regular interval. In this ous motion.
 CHAPTER 12  Digital Imaging 169

There is a direct relationship between the

speed with which changes are occurring and the
rate at which sampling is performed so that no
useful information is lost. The faster the signal
changes, the faster it must be sampled to describe
A all of the details of that change accurately with
the data stored in the computer.
Signals and image patterns that change rapidly
are described as containing high frequencies. Sig-
nals and image patterns that change slowly are
described as containing low frequencies.
Sampling must be rapid enough so that the
B signal being sampled does not change direction
more than once between samples. If the sampling
is not quick enough, a rapidly changing process
may appear to change slowly (Figure 12-6).
When a rapidly changing signal is undersam-
pled and appears to change slowly, the process is
said to be aliased because of the misleading
nature of the sampled data. The stagecoach in a
C
Wild West movie appears as though its wheels
FIGURE 12-6  A, A continuous free induction decay (FID). are moving backwards because the spokes appear
B, A sampled version of the FID in which the samples are
taken too far apart in time. C, The connect-the-dots rep- to rotate backwards (Figure 12-7). The movie
resentation of the undersampling shows that the higher frame rate (sampling rate) was not fast enough
frequency has been aliased into a lower frequency.

FIGURE 12-7  The apparent backward rotation of wagon wheels in Wild West movies is due to aliasing. The frame rate
is not fast enough to record the actual motion. Aliasing in magnetic resonance imaging results in a “wraparound”
artifact.
170 PART IV  Image Formation

FIGURE 12-9  This patient was positioned too far anteri-

orly in the head coil; the result was wraparound posteri-
FIGURE 12-8  This is a computed tomography image of
orly. (Courtesy Errol Candy, Dallas, TX.)
a plastic cylinder containing an air-filled triangular section.
The dark streaks appearing off the edges of the triangles
are the result of aliasing. (Courtesy Edward Nickoloff, New
York, NY.)
because the signal originates from various regions
of the body. Thus the frequencies to be observed
to capture the true motion of the wheels. The are those of spatial frequency.
result is an aliased image.
Aliasing results in a permanent loss of infor-
THE SPATIAL FREQUENCY
mation. An example of an aliasing artifact, not
DOMAIN
uncommon in CT, appears as streaks from a very
sharp interface (Figure 12-8). Aliasing artifacts Spatial resolution is measured in line pairs per
in MR images appear as “wraparound” struc- centimeter (lp/cm) or line pairs per millimeter
tures (Figure 12-9). MR artifacts are described (lp/mm). Spatial frequency is measured in cycles
more completely in Chapter 28. per unit distance (e.g., cycles/cm or cycles/mm).
Using a digital computer creates two opposing The two are nearly identical in meaning and
constraints. The limited amount of data storage intuition because line pairs per unit distance
capacity argues for taking as few samples as pos- applies to cycles per unit distance as well.
sible. The desire for high spatial resolution argues The ability to image high spatial frequencies
for taking as many samples as possible. is the ability to image very small objects. Five line
The lower limit on sampling is determined pair patterns and five spatial frequency patterns
by the data because enough samples must be are shown in Figure 12-10. The larger patterns
acquired to avoid aliasing. The upper limit is are represented by smaller numbers. As spatial
determined by the capacity of the computer. frequency increases, object size decreases and
MRI must also sample the information about becomes more difficult to image. The spatial
the patient. The number of samples in MRI is resolution of the best clinical MRI system is
limited by the time available to collect the data approximately 10 lp/cm or 1 lp/mm. Table 12-3
from a very short signal. The changes of the shows the approximate spatial resolution for
signal are changes in not only time but also space some other medical imaging modalities.
 CHAPTER 12  Digital Imaging 171

One line pair

Object

Spatial
frequency 1 2 3 4 5
Ip/mm

Image

FIGURE 12-11  Three entrepreneurs and their working

Image attire demonstrate the concept of spatial frequency. The
contrast 0.88 0.59 0.31 0.11 0.01
used car salesman’s plaid jacket contains high spatial fre-
FIGURE 12-10  These five line pair patterns are quencies both horizontally and vertically. The undertaker’s
spatial frequency patterns used to demonstrate spatial plain black jacket has zero spatial frequency. The banker’s
resolution. pinstripe suit has zero vertical spatial frequency but higher
horizontal spatial frequency.

TA B L E 1 2 - 3 Approximate Spatial Answer: 10 lp/cm = 20 objects/cm, therefore

Resolution Capability 1 object = 1/20 cm
of Several Medical = 0.05 cm
Imaging Modalities = 0.5 mm
Question: A 256 × 256 MR image is acquired
Spatial over a 10 cm field of view (FOV). What is the
Object Size Frequency limiting spatial frequency?
Imaging Modality (mm) (lp/mm)
100 mm
Radioisotope scan 10.0 0.05 Answer: pixel size = = 0.4 mm
Ultrasound 2.0 0.25 256
Computed tomography 0.5 1.0   2 pixels = 1 lp
Magnetic resonance 0.5 1.0   therefore 0.8 mm = 1 lp
imaging
Intensified fluoroscopy 0.15 3.0 1 lp
Screen-film 0.05 10
  and = 1.25 lp/mm
0.8 mm
Human eye 0.05 10.0
Direct exposure film 0.02 25.0 A high spatial frequency reflects many changes
in the intensity of the MR signal. This is a func-
tion of location within the patient. A low spatial
frequency arises from few or no changes in the
Space—Spatial Frequency
intensity of the MR signal across the patient.
 1  An example of different spatial frequencies
cm = 1/2 
 lp/cm  can be illustrated by the three businessmen
1  shown in Figure 12-11: a used car salesman, an
lp/cm = 1/2 
 cm  undertaker, and a banker. The used car salesman
is wearing a loud plaid jacket, the undertaker is
wearing a plain black suit, and the banker is
Question: An MRI system is capable of 10 lp/cm wearing a pinstripe suit.
resolution. What pixel size does this The pattern of the fabric of the used car sales-
represent? man’s plaid jacket has many abrupt changes in
172 PART IV  Image Formation

A B
FIGURE 12-12  A, The original image with a sampling grid superimposed. B, The sampled values of the original image.
Because of sampling in space, the resulting computer image may look blocky.

Column number
1 2 3 4 5 6 7 8 9 10
1
2 37 63 82 7 16
3
4 12 18 98 31 22
Row 5
number 6 41 8 39 73 3
7
8 14 48 66 53 49
9
A 10 B C 28 82 71 33 19

FIGURE 12-13  A, 10 × 10 matrix of cells. B, 5 × 5 matrix of cells. C, 5 × 5 matrix of numbers in imaginary cells.

any square centimeter. Thus there are many high

Image Matrix
spatial frequencies in that jacket because there
are many changes per centimeter, whether the Because the computer has a limited amount of
jacket is sampled horizontally across the cloth or memory in which an image can be stored, the
vertically. Whether the operator samples verti- image must be sampled in space as well as in
cally or horizontally, there are no changes in the time. For the sampling of an image in space, the
pattern of the fabric of the undertaker’s solid values are selected at the intersections of an
black suit, and thus the jacket has very low imaginary grid, which is superimposed on the
spatial frequencies. image (Figure 12-12), and the values represent a
Between these extremes is the banker’s pin- small region of the image.
stripe suit. A vertical sample on the back of the The term image matrix refers to a layout of
jacket reveals no change. Therefore the jacket has rows and columns, usually containing numbers
very low spatial frequencies vertically. On the representing intensity in boxes or cells. Figure
other hand, when the jacket is sampled horizon- 12-13 shows a 10 × 10 matrix of cells, a 5 × 5
tally across the fabric, the pinstripes are crossed, matrix of cells, and a 5 × 5 matrix of numbers
and some relatively high spatial frequencies are in imaginary cells. Each MR image consists of a
observed. The more closely spaced the pinstripes, matrix of imaginary cells, each having various
the higher the horizontal spatial frequency. brightness levels (the gray scale). The brightness
 CHAPTER 12  Digital Imaging 173

of a cell is determined by the computer-generated

number in that cell.
The size of the image matrix is determined by
characteristics of the imaging system and the
capacity of the computer. Most MRI systems
provide image matrix sizes of 256 × 256, although
often other sizes such as 192 × 256 and 512 ×
512 are available. Sometimes, an arbitrary acqui-
sition matrix, such as 384 × 512, is placed in the Original 128  128
center of the nearest larger power-of-2 matrix,
512 × 512 in this case, with the unused cells at
minimum intensity.
The spatial resolution of any digital image is
limited by pixel size. Therefore spatial resolution
is improved with a larger image matrix, assuming
that the FOV of the image is held constant.

Spatial Resolution 64  64 32  32

FOV
Pixel size =
Matrix size

Figure 12-14 illustrates the influence of

matrix size on image quality when the FOV
remains fixed. A 64 × 64 cell matrix appears
definitely “blocky,” whereas a 128 × 128 matrix
is a fairly good representation of the original 16  16 88
image.
In theory, the sampled value represents the
intensity of the original continuous image at an
infinitesimally small point. Such a point would
be invisible, so each point is represented by a
pixel of the displayed image.

Spatial resolution is limited by pixel size.

44 22
FIGURE 12-14  The small features and sharp edges in the
A sampled image is just like a mosaic. Each image that contain much of the high spatial frequency
pixel is a small tile having a brightness equal to information are progressively obscured as the sampling
that of the sampled value at that point in the rate and image matrix are reduced. The fine structures
become blurred as their high spatial frequencies are aliased
image. This accounts for the possible blocky into lower spatial frequencies.
appearance of some computer-displayed digital
images. apart, it would be difficult to say precisely where
The spacing of the samples must be sufficiently the structure should be located in the image
close so that there are no surprises between and the precise shape of the structure. Sharp
samples. For example, if there were a small struc- edges and small structures contain high spatial
ture in the patient and the samples were far frequencies.
174 PART IV  Image Formation

A B C
FIGURE 12-15  The spatial frequency representation of an image. A, The original image. B, The magnitude map of the
spatial frequencies. C, The phase map of the spatial frequencies. The image is predominately positive in intensity, and
the center of B is therefore very bright because the average value in the image is quite large.

vertical spatial frequency does not affect the

Frequency Domain Map
information itself but only how it is presented.
The spatial frequency content of an image is
displayed in two dimensions, just the way the
SPATIAL LOCALIZATION AND
image itself is displayed. Low frequencies are
MAGNETIC RESONANCE IMAGING
near the center of the spatial frequency display
and higher spatial frequencies are toward the Although the topic of this book is MRI, comput-
edges (Figure 12-15). ers, spatial frequency, and sampling are empha-
Horizontal spatial frequencies appear along sized because MRI samples the strengths of the
horizontal lines, and vertical spatial frequencies spatial frequencies of the MR signal in the patient
appear along vertical lines. The intensity of an to get the image information. The computer then
arbitrary point represents the strength of a par- converts that information from its spatial fre-
ticular combination of horizontal and vertical quency representation to a representation in the
spatial frequencies. For example, the banker’s spatial domain or position. The latter is the usual
pinstripe suit was a combination of low vertical definition of an image, but the former is how the
spatial frequency and higher horizontal spatial raw data are presented to the computer.
frequency. Several MRI techniques have been developed.
The information in the spatial frequency rep- However, the two-dimensional Fourier trans-
resentation of the image is exactly the same as form (2DFT) method has led people to concen-
the information in the image; it is just being trate on the spatial frequency domain, or
represented in a different way. There are exam- “k-space” (see Chapter 13). What once seemed
ples of this concept in everyday life. Scheduled to be very different methods of imaging, projec-
patients appear on the MRI system worksheets tion reconstruction and 2DFT, are just different
in chronological order. However, when they are methods of measuring the spatial frequency
billed for the procedure, the list is likely to be in domain. They have been joined by some other
alphabetical order or arranged by hospital methods, all of which are easily understood in
number. The same information is present; a the spatial frequency domain.
group of patients received MRI examinations.
The manner in which the data are presented is
The spatial frequency information of the patient
different for different purposes. Whether the
is sampled into the spatial frequency domain and
information in an image is organized by horizon-
ordered in the computer as k-space.
tal and vertical position or by horizontal and
 CHAPTER 12  Digital Imaging 175

Projection reconstruction, 2DFT, and three-

dimensional Fourier transform (3DFT) MRIs are
digital imaging techniques. They differ in how
they sample the spatial frequencies of the patient.
The principles of spatial localization demonstrate
why MRI is inherently digital, regardless of the
particular technique used.
In a homogenous magnetic field, all of the
spins precess at exactly the same rate. This rate
equals the Larmor frequency. The Larmor equa- Gradient magnetic field BZ
tion asserts that the frequency of precession is
directly proportional to the strength of the mag- FIGURE 12-16  The effect of a gradient magnetic field,
netic field, B0. Because the frequency, phase, and BZ, on the resonant frequency of spins within the patient.
amplitude of the signal are all that can be mea- Note that the spins in the head precess more slowly than
those in the belly.
sured, it becomes necessary to modify the mag-
netic field so that the MR signal can reflect its
spatial origin. the spatial frequency representation of the image
slice makes MRI inherently digital imaging.
The difference between a nuclear magnetic reso-
nance spectrometer and an MRI system is the
Projection Reconstruction Magnetic
presence of gradient magnetic fields in MRI.
Resonance Imaging
Gradient magnetic fields are used to achieve The projection reconstruction MRI technique is
spatial localization. When the strength of the the easiest to understand because it resembles the
magnetic field varies linearly across the patient, way in which data are collected in parallel beam,
the frequency of the signal does also. Spins on one x-ray CT. As two orthogonal gradient magnetic
side of the patient precess faster than spins on the fields are applied simultaneously, the direction of
other side. When the MR signal is received, the the resulting gradient is the vector sum of these
range of frequencies in the patient will be detected two orthogonal gradients. The direction of the
and clearly distinguished. As shown in Figure resultant vector sum gradient may be rotated in
12-16, the MR signal from the patient’s head is of the imaging slice by varying the relative strengths
lower frequency than that from the abdomen. of the two orthogonal gradient magnetic fields,
Gradient magnetic fields also provide the BX and BY (Figure 12-17).
means to distinguish left from right within the A separate MR signal is acquired for each
slice of the image but do not yield any informa- orientation of the resultant vector sum gradient.
tion about the distribution of intensity along the This MR signal is a sample of the two-dimensional
direction at right angles to the gradient. For spatial frequency representation of the image. As
information to be obtained throughout the plane the resultant vector sum gradient is rotated from
of the object, two gradient magnetic fields must acquisition to acquisition, the orientation of the
be used. samples in the spatial frequency domain rotates.
The two gradients are applied simultaneously Thus the spatial frequencies in the image are
in projection reconstruction imaging and sequen- sampled on a polar grid (Figure 12-18).
tially in 2DFT imaging. These approaches to Back projection is a time-consuming compu-
MRI reflect different ways of coordinating the tational chore that tends to emphasize the noise
effect of the two orthogonal gradient magnetic in the projections. For these reasons, the 2DFT
fields. The combination of the two gradients and 3DFT techniques have become the methods
allows the whole slice to be sampled. Sampling of choice for MRI.
176 PART IV  Image Formation

BY
BXY

Projection no. 1

BX

Projection no. 2

Projection ... etc.

FIGURE 12-17  The vector sum of two simultaneously applied gradient magnetic fields can be rotated by varying the
strength of each gradient. Note that the strengths of the X and Y gradients are chosen so that their strength always
remains the same, even though the angle of the resultant BXY changes.

Two-Dimensional Fourier Transform

Magnetic Resonance Imaging
The 2DFT approach also samples the spatial fre-
quency domain of the image but does so on a
rectangular grid instead of a polar grid (Figure
12-19). The 2DFT technique consists of a basic
cycle that is repeated many times, typically 256.
This cycle consists of radiofrequency (RF) excita-
tion pulse (RFt), followed by a magnetic field
pulse from the phase-encoding gradient (BΦ) and
then by a steady application of an orthogonal
gradient called the read or frequency-encoding
gradient magnetic field (BR), during which time
the MR signal is detected.

FIGURE 12-18  Polar sampling of the spatial frequency From one signal acquisition to the next, only the
domain of the image. Note that the angle of the resultant strength of the phase-encoding gradient mag-
BXY gradient determines the angle of the sample line in netic field is changed.
the spatial frequency domain.
 CHAPTER 12  Digital Imaging 177

happens, the image is aliased, and the bottom

of the image appears to wrap around the top
(Figure 12-20).
The cure for aliasing is to increase the number
of samples in the data collection stage or increase
the FOV. Both of these solutions increase the
sampling frequency.
The effects of the subject’s motion on the
2DFT image are complicated. Most of the visible
FIGURE 12-19  Rectangular sampling of the spatial fre- effects are in the phase-encoded direction of the
quency domain. The phase-encoding gradient is changed image. This is because a small amount of motion
during each signal acquisition, which determines the hori- will cause small frequency changes but large
zontal line of the spatial frequency domain.
phase changes.
The effects of motion are determined by the
After a suitable delay, the cycle is repeated. size of the displacements relative to the data sam-
The phase-encoding gradient magnetic field pling times. The time between the successive
selects a single line in the k-space representation applications of the phase-encoding gradient is
of the image. Then the frequency-encoding gradi- hundreds to thousands of milliseconds, repeti-
ent magnetic field forms the MR signal along this tion time (TR), whereas the time between sam-
line. Instead of rotating the sampling line around pling data points along the frequency-encoding
the center of k-space, the phase-encoding gradi- gradient, BR, is microseconds.
ent magnetic field shifts the MR signal so that it The motion effect is sometimes useful in iden-
samples a different line of spatial frequencies tifying the phase-encoding direction by noting
parallel to the others. When the strength of the the direction of signal ghosting. This is important
phase-encoding gradient magnetic field is changed to know because motion can still produce arti-
in following cycles, other lines in k-space are facts even when the image is cardiac gated and
measured. respiratory gated (Figure 12-21). Through judi-
A family of lines in k-space has been selected, cious selection of the phase-encoding direction,
and the frequency information along those lines the overlap of the artifacts and organs of interest
has been measured. This rectangular sampling is can be minimized.
Fourier transformed to yield the image. Because
the k-space data are already sampled on a rect-
Three-Dimensional Fourier
angular grid, 2DFT replaces the filtering and
Transform Magnetic
back projection steps of projection reconstruc-
Resonance Imaging
tion with a second Fourier transform.
Bigger and faster computers have made it possi-
The two most common artifacts peculiar to 2DFT ble to store contiguous 2DFT slices and display
images are the consequences of undersampling them as an apparent three-dimensional image. A
and motion. faster way of acquiring data for three-dimensional
display is to follow the second Fourier transform,
Undersampling means that not enough points which produces a slice image, with yet another
are used in the data collection. This results in the Fourier transform. With this method, a three-
samples not being sufficiently close in the fre- dimensional image is obtained more quickly and
quency representation of the data. When this with improved spatial resolution.
178 PART IV  Image Formation

FIGURE 12-20  Aliasing in two-dimensional Fourier transform images can result in “wraparound” of tissues in the
patient. (Courtesy R. Mark Henkelman, Toronto, Canada.)

A B C
FIGURE 12-21  The effects of motion on two-dimensional Fourier transform (2DFT) images. A, An ungated cardiac
image. B, End diastolic image of a heart. C, End systolic image of a heart (B and C are of the same heart). The ungated
image has extremely poor definition of the myocardium and prominent streaking artifacts in the phase-encoded direction.
The gated images show improvement in the motion artifacts, but these artifacts are not eliminated because the motion,
even when gated, causes phase errors that mislead the 2DFT reconstruction algorithm.
 CHAPTER 12  Digital Imaging 179

7. The patient aperture for most MRI gantries

CHALLENGE QUESTIONS
is approximately 60 cm. Express this in
1. What is the best spatial resolution possible binary form.
with any digital image? 8. An MRI examination results in 32 images of
2. What is the difference between a digit and a 256 matrix size and 12-bit gray scale. How
bit? much disk space will be required to store
3. When acquiring data for an MR image, these images, uncompressed?
what must change from one signal acquisi- 9. Explain why a 12-bit dynamic range is useful
tion to the next? for MRI when the dynamic range of the
4. What would one have to do to turn a nuclear human eye does not exceed 5 bits.
magnetic resonance (NMR) spectrometer 10. As spatial frequency increases, what happens
into an MRI system? to object size, spatial resolution, and the
5. An MRI system is set to have a 12-bit contrast resolution?
dynamic range. How many gray levels can
that imaging system display?
6. The best resolution MRI systems can obtain
is approximately 2 lp/cm. What is the equiv-
alent spatial resolution expressed in size?
 CHAPTER

13 
A Walk Through the
Spatial Frequency Domain

OBJECTIVES
At the completion of this chapter, the student • Draw several graphic representations of a
should be able to do the following: Fourier transform.
• Define spatial frequency, spatial frequency • Identify a sinc pulse and its appearance
domain, and k-space. graphically.
• Relate the forward and inverse Fourier • Describe the relationships among receiver
transforms to their place in two-dimensional bandwidth, signal-to-noise ratio, and field
Fourier transform (2DFT) magnetic resonance of view.
imaging (MRI).

OUTLINE
Spatial Frequency Signal-to-Noise Ratio
The Fourier Transform Contrast
Impact of Spatial Frequencies Artifacts
Spatial Resolution Spatial Frequency Patterns and
Field of View Order

KEY TERMS
K-space Signal-to-noise-ratio Temporal frequency
Line pair Spatial frequency

The spatial frequency domain is crucial to mag- recorder and is used to listen to a musical ensem-
netic resonance imaging (MRI) because all but ble, the result is a tracing of sound intensity as a
the very earliest methods of MRI (such as the function of time; the microphone simply responds
sensitive point method) scan the spatial frequency to changes in air pressure at its sensor.
information about the image, not the image When a member of the audience hears
plane itself. the same performance, the individual instru-
A useful analogy is the human auditory system. ments of the ensemble are perceived as they play
When a microphone is attached to a strip chart in concert (Figure 13-1). This is because the

180
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 181

A B
FIGURE 13-1  A, The tracing of the output of a microphone picking up the performance of an ensemble. The instru-
ments in the ensemble produce a combined sound that is recorded as a single signal. B, A listener enjoys the performance
of the ensemble and hears each instrument individually. This is because of the human auditory system’s ability to perform
a real-time frequency analysis.

human auditory system converts the variations One line pair

of loudness observed by the microphone into a
temporal frequency representation from which Object
the brain extracts higher-level information about
the instrumentation. Spatial
1 2 3 4 5
In MRI, the imaging system measures the frequency (lp/cm)
information about the image slice differently
than the human visual system would if the slice
Image
were directly visible. A method of analyzing the
spatial frequency components and converting
Image fidelity 0.88 0.59 0.31 0.11 0.01
them into spatial location information is needed
to reconstruct the image. The usual method is the FIGURE 13-2  The fidelity of the image of a bar pattern
decreases with increasing spatial frequency. This loss of
Fourier transform (FT). fidelity of a high-contrast object results in a low-contrast
The purpose of this chapter is to extend the image.
introduction to spatial frequency in the preced-
ing chapter and to discuss the FT. Chapter 14
demonstrates how the MRI system actually mea-
sures spatial frequencies. Spatial frequency has units of line pair/cm or line
pair/mm—10 lp/mm = 1 lp/cm.

SPATIAL FREQUENCY Consider a series of high-contrast bar patterns

The term spatial frequency domain comes from to be imaged (Figure 13-2). One bar and its equal
the field of electrical engineering. The term width interspace are called a line pair.
k-space comes from physics. They refer to the The number of line pairs per unit length is the
same thing. Either is correct and both are in spatial frequency, and for MRI systems it is
common use. The former is far more descriptive; expressed in line pair per centimeter (lp/cm). A
however, k-space is firmly entrenched in MRI low spatial frequency represents large objects,
and therefore is used here to emphasize the and a high spatial frequency represents small
subject under discussion. objects.
182 PART IV  Image Formation

1.0
0.9
0.8
0.7 5
4
0.6
MTF 0.5
53.1
0.4
0.3 3
0.2
0.1
0
1 2 3 4 5 6
Spatial frequency (lp/cm)
FIGURE 13-3  Modulation transfer function (MTF) is a
plot of the image fidelity versus spatial frequency. The  FIGURE 13-4  A net magnetization vector in the XY
five data points plotted here are from the analysis of  plane, as represented in polar coordinates.
Figure 13-2.

The image obtained from the low-frequency less tangible about one part than the other, at
bar pattern more faithfully reproduces the least in the MRI context. This is simply a math-
object than that of the high spatial frequency ematical convenience.
bar pattern. The less faithful reproduction with Complex numbers are important in MRI. For
increasing spatial frequency describes the limita- example, as the net magnetization vector pre-
tion of spatial resolution of an imaging system cesses in the XY plane, it takes two numbers to
and is indicated graphically as a modulation specify where it points at any given time. One
transfer function (MTF). number gives the X coordinate and the other
A graph of the ratio of image-to-object at the number gives the Y coordinate. Thus if the X-axis
spatial frequencies in Figure 13-2 results in the value is assigned to the real part of a complex
MTF curve of Figure 13-3. For MRI, the MTF number and the Y-axis value to the imaginary
is obtained from the FT of an edge or impulse. part of a complex number, the position of the net
The 10% MTF value is usually stated as the magnetization vector can be represented by a
limiting spatial resolution of an MRI system, single complex number.
and that value depends on matrix size and field Two coordinate systems are commonly taught
of view (FOV). Limiting spatial resolution is in geometry. One is the Cartesian coordinate
approximately 10 lp/cm for head/body images system, in which a point in a plane is represented
and 20 lp/cm for surface coil imaging. by its X and Y coordinates. The other system is
A spatial frequency is a number of cycles in polar coordinates, in which a point is represented
space or per unit distance. Cycles of what? The by its distance from the origin and the angle that
answer is a bit tricky. In the previous chapter, the the line connecting it to the origin makes with
complex nature of the magnetic resonance (MR) the X-axis (Figure 13-4). Polar coordinates fit
image was not emphasized, but now it is very MRI naturally.
important. When the net magnetization vector precesses
A complex number is one that has two dimen- in the XY plane, its length is essentially constant
sions, such as north-south and east-west or when compared with its rotation (change in
azimuth and range. The two dimensions are angular orientation) in the XY plane. If polar
called the real part and the imaginary part. These coordinates are used to describe the position of
names are unfortunate; there is nothing more or the net magnetization vector in the XY plane, the
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 183

Y frequencies in an object can be measured, the

object can be described. In a formal sense, that
X is usually done by the FT.
During imaging, each MR signal is Fourier
X
transformed to fill a single line of k-space
(Figure 13-7). High-amplitude phase-encoding
Y gradients fill the periphery of k-space and con-
tribute to spatial resolution. Low-amplitude
phase-encoding gradients fill the center of k-space
Magnitude and contribute to contrast resolution.
The low-amplitude phase-encoding gradients
result in high-intensity MR signals. The zero
Phase phase-encoding gradient that fills the central line
FIGURE 13-5  The XY plane component of the net mag- of k-space produces the highest-intensity signal
netization vector of each voxel is plotted. The orientation line of k-space.
of the net magnetization vector describes a spiral in space. When k-space is filled, there exists a matrix
The pitch of the spiral is proportional to the spatial
array of cells, each containing a real and an
frequency.
imaginary number. These two number arrays
result in the magnitude and phase maps present
before the second FT.
length of the vector is constant but the angle The k-space matrix data features are related
changes as the vector precesses. Temporal fre- to the reciprocal of the corresponding image
quency is measured in cycles per second (hertz space features. For instance, the size of the image
[Hz]). One cycle equals 360°. space field of view (FOV) is related to one divided
It is important to be clear about whether you by the time between MR signal samples. The
are talking about frequency or angular frequency. bigger the size of k-space sampled, the smaller
Angular frequency has units of radians per the structures that can be resolved.
second (360° = 2π radians = 6.28 radians). Often
the gyromagnetic ratio, γ, is expressed in radians There is an inverse relationship between the
and calculations involving it can be off by a matrix of k-space and the image.
factor of six if you are not careful.
It is relatively easy to imagine the net magne- Each matrix cell in k-space contains informa-
tization vector rotating as a function of time as tion about all of the image pixels. Also, each
it precesses. If the net magnetization vectors posi- image pixel contains information obtained from
tioned along the patient have different angles, each and every matrix cell. This is important to
that represents a spatial frequency (Figure 13-5). understand but difficult to conceptualize.
Any object can be viewed as a weighted sum Going back to our analogy regarding patient
of spatial frequencies. There are mathematical records, often the data are stored files that
conditions under which this statement is not contain all the data on each patient. Alterna-
true, but any patient who met those conditions tively, you could have all of the patient Zip Codes
would have an acute need for treatment far in in one file, all patient telephone numbers in
excess of an MRI. another file, and so on. The utility of arranging
A few examples of real and imaginary FT the information in this manner will depend on
pairs are shown in Figure 13-6. Notice the simi- what you want to do with it. For example, if you
larity between these images and Figures 8-10 want to know the geographic distribution of
through 8-13. The important thing to notice patients served by the hospital, the Zip Code
about these examples is that if the spatial data would be all you need, and having to go
184 PART IV  Image Formation

C
FIGURE 13-6  Several examples of real and imaginary Fourier transform pairs. A, 8 horizontal cycles. B, 16 vertical cycles.
C, 16 cycles + 18° phase shift. (Courtesy Bud Wendt, Houston, TX.)

KY

FT KX

Bφ SE or GRE
FIGURE 13-7  During spin echo (SE) magnetic resonance imaging, each SE is sampled and Fourier transformed to fill
one line in k-space. FT, Fourier transform; GRE, gradient echo.
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 185

180
FT 90
α
single frequency single frequency

FT
gaussian gaussian
FIGURE 13-9  The magnetic resonance imaging broad-
band sinc pulse is symbolized as shown for the three types
of RFt.

FT
why the RFt shape in Figure 13-9 is used through-
square wave sinc function out this text.
FIGURE 13-8  When the signal or mathematical function The FT (sometimes called the forward trans-
on the left is Fourier transformed, the result is to the right. form to distinguish it from the inverse transform)
The Fourier transform (FT) of the right is the inverse Fourier is involved mathematically but is simple concep-
transform (FT−1) and results in the function to the left. tually. The FT compares the object with a set of
test functions and reports how similar the object
is to each test function. When combined, the test
through patient files to get all the Zip Codes functions are defined by the FT and have the
would be very tedious. property that they can completely describe all
possible objects.
The forward transform is used to build a set
THE FOURIER TRANSFORM of functions that contain all of the information
The FT is a mathematical method for analyzing that was in an image, but which is now repre-
the frequency content of something. In the sented in terms of spatial frequencies. The inverse
current discussion, it applies to the spatial fre- FT constructs the image from a set of spatial
quency content of a patient. frequency components. Since the MRI signal
The inverse FT analyzes the set of spatial contains the information in terms of spatial fre-
frequency values given to it and indicates the quencies, the inverse transform is used to process
object corresponding to those frequencies. Tech- MRI raw data.
nically, the inverse FT is used to reconstruct Rather than enumerating the number of
an MR image; that is, to convert the spatial fre- objects in an image, the forward and inverse FT
quency information measured by the imaging functions perform comparisons to determine the
system into an image. exact details of the spatial relationships between
The FT converts a time-varying signal into its the structures depicted in an image. Each of the
frequency components. Figure 13-8 shows this test functions represents a different spatial fre-
relationship for several functions important to quency. The two parameters determined by the
MRI. Note the two lower functions. The FT comparison are the magnitude—that is, how
of a rectangular function is an oscillating fre- much of that spatial frequency is contained in the
quency spectrum called a sinc function. The FT object—and the phase, or what is the starting
of a sinc function is a band of frequencies of angle for that spatial frequency.
equal intensity.
This sinc function is of particular importance
The Fourier transform (FT) relates the number and
to MRI because its FT is rectangular, as in a
positions of spatial features to the distances
rectangular slice. Most MRI radiofrequency (RF)
between and the frequency of their occurance.
pulses are sinc functions for this reason. That is
186 PART IV  Image Formation

A B C
FIGURE 13-10  256, 128, and 64 phase-encoding step acquisitions of a transverse image of the head.

The inverse transform takes the magnitude

Field of View
and phase information for each spatial frequency
and, by knowing what the appearance of the test Generally, FOV is an operator-defined parameter
function associated with those data looks like, that controls the apparent size of the patient in
reconstructs the object. Because the MRI raw the image. The matrix size is the operator-defined
data are the magnitudes and phases of the spatial parameter that, with the FOV, controls the pixel
frequencies, the inverse FT is used to reconstruct size in the image.
the image.
Pixel Size
IMPACT OF SPATIAL FREQUENCIES Field of view
Pixel size =
Matrix size
Spatial Resolution
The basic rule of thumb is that higher spatial The pixel size is the dimension of the FOV
frequencies result in better spatial resolution. divided by the number of points in the acquisi-
Thus if a sharp, crisp image is desired, it is neces- tion matrix in that direction. It takes two pixels
sary to measure not only the low spatial frequen- to define a line pair, so the resolution is one line
cies but high spatial frequencies as well. For pair per the width of two pixels, and the highest
example, if two objects separated by 1 mm need spatial frequency is one cycle per width of two
to be resolved, a resolution of 1 line pair per pixels.
millimeter (lp/mm) is necessary. This implies that
the spatial frequencies must be measured to at Question: A 256 (wide) × 128 (high) acquisition
least 1 cycle/mm. matrix has a 25.6 cm FOV. What is the
Figure 13-10 illustrates the same image spatial resolution and maximum spatial fre-
measured with different numbers of spatial fre- quency in each direction?
quencies. The importance of the higher spatial Answer: Horizontally, each pixel is 256 mm/256
frequencies to spatial resolution is obvious. The pixels = 1 mm wide. Thus there is 1 line pair
way to improve the spatial resolution is not per 2 mm or 0.5 lp/mm horizontal resolu-
to increase the density of the measurements of tion, and the maximum unaliased spatial fre-
spatial frequency, but rather, to measure higher quency is 0.5 cycle/mm. Vertically, each pixel
spatial frequencies. is 256 mm/128 pixels = 2 mm high. Thus
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 187

Low-frequency signal

Noise

High-frequency signal

Noise

FIGURE 13-11  Signal-to-noise ratio is usually higher with low-frequency signals.

there is 1 line pair per 4 mm or 0.25 lp/mm High SNR improves contrast resolution.
vertical resolution, and the maximum una-
liased spatial frequency is 0.25 cycle/mm.
This is not to say that high spatial frequencies
The image in this example may be interpolated are unimportant visually. Sometimes small struc-
to a 256 × 256 matrix for viewing, but that does tures and fine detail are the reasons to acquire
not alter the actual resolution, which is deter- the image. Often mathematical importance and
mined by the data acquisition. clinical importance are two separate issues. It
should be kept in mind, however, that the higher
spatial frequencies are typically more noisy, and
Signal-to-Noise Ratio thus improving the spatial resolution of the
The signal-to-noise ratio (SNR) is a comparison image results in a noisier looking image.
of the intensity of the information (signal) in the Usually the radiologist makes the determina-
image to the intensity of the noise in the image. tion of the relative importance of resolution and
Most images have more signal strength at low SNR for each protocol. Some types of images
spatial frequencies and less signal strength at have more inherent SNR than others, therefore
high spatial frequencies. Most noise is uniformly allowing more flexibility in trading off spatial
distributed in spatial frequency. Thus, if SNR is resolution and noise.
considered as a function of spatial frequency, it
is much better at low spatial frequencies than at Narrow receiver bandwidth results in better SNR.
high spatial frequencies (Figure 13-11).
188 PART IV  Image Formation

Center frequency

Noise

16
32
Frequency bandwidth (kHz)
FIGURE 13-12  Narrow receiver bandwidth results in higher signal-to-noise ratio.

The term bandwidth refers to the range of

Contrast
frequencies contained in an RF pulse. There are
two RF bandwidths of importance to MRI: Contrast is a property of the spatial domain. For
transmitted (RFt) and received (RFs). The RFt is example, two pixels, one of which has twice the
involved in slice thickness and image weighting brightness of the other, have a contrast ratio of
and does refer to a range of frequencies. 2 : 1. If the pixels are far apart, they represent a
Receiver bandwidth is principally involved low spatial frequency. If they are adjacent, they
with SNR and FOV. The receiver bandwidth represent a high spatial frequency.
refers to the sampling rate of the MR signal Thus the relationship between contrast and
(RFs). The receiver bandwidth is simply the spatial frequency depends on the particular
number of data points sampled (N) divided by image. However, most of the amplitude in k-
the total signal digitization time (T). Put another space is concentrated in the low spatial frequen-
way, the receiver bandwidth is inversly propor- cies for most typical clinical imaging situations
tional to the time between data point samples (with MR angiography being one notable excep-
(Δt) (RFS = N/T = 1/Δt). tion), and thus the contrast in an image lies in
The SNR is higher with narrow bandwidth the low spatial frequencies.
than with wide bandwidth. This is because
noise is nearly uniform regardless of frequency
Artifacts
and is therefore called white noise (Figure 13-12).
A signal with narrow bandwidth contains MRI artifacts result from a variety of character-
relatively less noise than a signal with wide istics of the imaging system and the imaging
bandwidth. process (see Chapter 28). However, there is one
At a fixed gradient magnetic field, receiver example, sometimes called a spark artifact, that
bandwidth determines FOV. If the read gradient actually helps to illustrate the points of this
magnetic field (BR) is constant, a narrow band- chapter. The spark artifact is a spike of noise,
width will result in a proportionally reduced perhaps from the discharge of a static buildup
FOV and a proportionally reduced pixel size. For in the imaging system, which appears as an iso-
a fixed FOV the read gradient magnetic field is lated bright spot in k-space because it was a
lower, the receiver bandwidth is lower, and the momentary occurrence during the sampling of
resulting SNR is higher (Figure 13-13). the spatial frequency domain. This bright spot
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 189

High bandwidth Low bandwidth

noise

A Strong gradient (BR) B Weak gradient (BR)

FIGURE 13-13  A, At a fixed gradient magnetic field, receiver bandwidth determines field of view (FOV). B, At a fixed
FOV, signal-to-noise ratio improves with lower receiver bandwidth.

A B C
FIGURE 13-14  Inverse transform of a normal image contaminated with a hot spot. A, Image. B, Magnitude of Fourier
transform (FT). C, Phase of FT.

will be treated as normal data by the reconstruc- Answer: Count the number of cycles of the arti-
tion algorithm, resulting in a striped pattern fact horizontally and vertically and divide by
(Figure 13-14). the FOV to get the spatial frequencies of KX
and KY. This is a magnitude image, and there-
Question: The FOV of the image in Figure 13-14 fore there is not enough information to know
is 20 cm. What are the horizontal and verti- whether the spatial frequencies are positive or
cal spatial frequencies of the spark artifact? negative.
190 PART IV  Image Formation

KX KX
Finish Finish

KY KY

Start Start
A Standard 2DFT B Blipped EPI
FIGURE 13-15  A, Diagram of the path through k-space taken by two-dimensional Fourier transform (2DFT) imaging.
The bottom row of the spatial frequencies is sampled left-to-right, the next row is sampled left-to-right, and so on until
all of the rows of k-space have been filled. B, A similar diagram for the blipped echo planar imaging (EPI) method. Note
the direction of sampling alternates left-to-right then right-to-left, thereby eliminating the need to jump back to the left
edge for sampling of each row.

SPATIAL FREQUENCY PATTERNS

AND ORDER
The spatial frequency domain representation of
any given object exists, and a variety of strategies
have been developed for measuring it. The sam-
pling concepts mentioned in Chapter 12 apply
here. It is necessary to make measurements of
the spatial frequency information sufficiently
close together so that there are no surprises. FIGURE 13-16  The projection reconstruction measure-
ments of spatial frequencies. The measurements are along
The easiest way to do this is with the diameters of a circle. The scan lines pass through the origin
rectangular raster image (Figure 13-15). This and are equally spaced in angle.
method is used for standard (2DFT) imaging (see
Chapter 14).
The measurement points are uniformly dis-
tributed in k-space. The trajectory in k-space is Projection reconstruction imaging was the
sometimes used to describe the locations of the first method of MRI and is still used in a few
measurements in the spatial frequency domain special applications, typically those requiring
because each sample point is measured at a extremely short echo times. The sampling scheme
slightly different time. The location of each mea- for projection reconstruction is shown in Figure
surement is controlled by the gradient magnetic 13-16.
fields. The density of measurements in projection
A slightly different path through k-space is reconstruction is greater in the center of the
used by the blipped echo-planar imaging tech- spatial frequency domain, that is, at the lowest
nique. Instead of scanning every line from left to spatial frequencies. This is the reason filtering is
right as in 2DFT, blipped echo-planar alternates used in the back projection method of recon-
the direction of the line scanning so that the structing these data. It compensates for the higher
pattern is a squared-off zigzag. density of samples at low spatial frequencies by
 CHAPTER 13  A Walk Through the Spatial Frequency Domain 191

A B C
FIGURE 13-17  Diagrams of the spatial frequency sampling paths of A, the spiral sampling method; B, the square spiral
sampling method; and C, the interleaved spiral sampling method.

attenuating the low spatial frequencies so that sampled in each of several acquisitions that are
the effective density is uniform. then combined into a complete measurement of
Another method of reconstructing projection k-space. The idea of this approach is to sample
data is to resample the spatial frequency mea- as many rows of k-space as time permits, for
surements onto a rectangular grid. They look like example, the end-systolic period of the cardiac
2DFT data and can be reconstructed by an cycle, and then to measure some more rows of
inverse FT. spatial frequency the next time conditions are
There are a number of relatively exotic appropriate.
methods of sampling k-space. These include This chapter has covered only the basics of
spiral scanning, square spiral scanning, and sampling the k-space. Researchers are discover-
interleaved spiral scanning (Figure 13-17). In ing more imaginative and productive ways of
general, these data are resampled onto a rectan- sampling k-space information in the imaging
gular grid and then reconstructed with the inverse slice or volume. Chapter 14 discusses the way
FT. The actual reconstruction algorithms are MRI actually measures spatial frequency.
complicated because of a variety of technical
factors including correcting for residual inhomo-
CHALLENGE QUESTIONS
geneities of the magnetic field.
The order in which the spatial frequencies 1. What is the measure of spatial frequency in
are sampled is a parameter over which the opera- MRI?
tor has some control in many MRI system soft- 2. MR image data are stored mathematically in
ware releases. Ordinarily, k-space is sampled something called k-space. Explain another
from bottom to top (see Figure 13-15). There term in which it is described.
is a class of MR data acquisition protocols, 3. What is the approximate limiting spatial
however, in which the relative contrast among resolution expressed in spatial frequency for
organs changes during the data acquisition. In an MRI system?
these images, it is usual to sample from the 4. What is meant by trajectory through k-space?
middle of k-space outward. This is based on the 5. A picture of k-space shows 256 lines. What
premise that the low spatial frequencies contain does that tell you about the image and the
most of the contrast information and should be imaging process?
measured early while the contrast among organs 6. As a general rule, when comparing a 256
is greatest. matrix image with a 512 matrix image, what
Another type of protocol is referred to as seg- can be said regarding spatial resolution and
mented k-space. Only a portion of k-space is contrast resolution?
192 PART IV  Image Formation

7. Which region of k-space, if any, is most 9. Describe the relationship of signal and of
important to the final image? noise across a range of spatial frequencies in
8. Because k-space is filled digitally, what is the MRI.
relationship between one pixel of k-space 10. How does a receiver bandwidth affect SNR
and one pixel of the image? and therefore contrast resolution?
 CHAPTER

14 
The Musical Score

OBJECTIVES
At the completion of this chapter, the student • Describe how a pixel is located within a slice.
should be able to do the following: • Relate two ways that slice thickness can be
• Identify the five lines of a magnetic resonance changed.
imaging (MRI) pulse sequence. • Draw the MRI pulse sequence diagrams for
• Recognize the various symbols used to partial saturation, inversion recovery, and
diagram an MRI pulse sequence. spin echo.
• Describe how a slice is selected for imaging.

OUTLINE
The Purpose of the Pulse Pixel Location within a Slice Pulse Sequence Diagrams
Sequences The Effect of a Gradient on Partial Saturation
Gradient Coil Precession Inversion Recovery
Function Phase Encoding Spin Echo
Slice Selection Frequency Encoding

KEY TERMS
Multi-echo imaging Pulse sequence Spin echo
Multislice imaging Signal averaging Spin warp

The magnetic resonance imaging (MRI) system its own line of music in the conductor’s score,
operates like a player piano. The control program each component of the MRI system (i.e., the
for the MRI system is a pulse sequence. The pulse transmitted radiofrequency [RFt], the received
sequence diagram is a complicated schematic MRI signal [RFs], and the three gradient mag­
containing graphs that show what each major netic fields coils, BSS, BΦ, BR) has its own row of
component of the MRI system should be doing timing information in the pulse sequence diagram.
at each moment. Fortunately, routine imaging uses prepro­
The pulse sequence diagram is analogous to grammed pulse sequences, so the MRI tech­
the musical score used by a conductor to lead nologist needs only to select a programmed
an orchestra. Just as each instrument follows technique.

193
194 PART IV  Image Formation

A B
FIGURE 14-1  These images of the same slice in the same patient illustrate that the timing of radiofrequency pulses not
only controls contrast but also can reverse contrast. A, A T1-weighted inversion recovery image. B, A T2-weighted spin
echo image.

THE PURPOSE OF THE The MRI pulse sequence diagram details the
PULSE SEQUENCES timing pattern for the RF pulses and gradient
After the preceding discussions of the physics of magnetic fields.
MRI and the necessary equipment for MRI, dis­
cussion can turn to how magnetic resonance In computed tomography (CT), the image
(MR) signals (RFs) make an image and what time, slice thickness, kVp, field of view,
characterizes the image. matrix size, and the reconstruction algorithm are
In MRI, a pulse sequence is a set of instruc­ selected by the technologist. These choices influ­
tions given to the imaging system to tell it how ence the contrast and spatial resolution of the
to make an image. In radiography, instructions image.
are given to the x-ray imaging system by setting MRI pulse sequences are analogous to
the kilovolt peak (kVp) and the milliampere- the radiographic and CT selections made by
second (mAs). The manner in which these the technologist. The pulse sequences specify the
controls, especially the kilovolt peak control, timing and magnitude of the RF pulses and the
are positioned influences the contrast of the gradient magnetic fields. The amplitude and
resulting image. timing of the RF pulses affect the image contrast.
In x-ray imaging, the relative order of the The pulsed gradient magnetic fields provide
gray scale remains unchanged. Regardless of the spatial localization and influence the spatial reso­
kVp, bone always appears brighter than soft lution of the image.
tissue. However, with MRI, the relative bright­ As in CT, the MR image is formed digitally.
ness of different tissues and the contrast rendi­ In conventional radiography, the image is analog.
tion can be reversed by changing the timing and The MR image is a mosaic of pixels. Each pixel
magnitude of the radiofrequency (RF) pulses has two properties: character and position. Char­
(Figure 14-1). acter refers to the intensity of the pixel.
 CHAPTER 14  The Musical Score 195

µ µw MRI signal
HU = K intensity (RFs)
µw

FIGURE 14-3  During proton relaxation, a complicated

magnetic resonance signal is received and computer pro-
CT MRI cessed to form an image.
FIGURE 14-2  Pixel brightness in CT is determined by the
x-ray attenuation coefficient, and in MRI, it is determined
by T1, T2, and proton density. the magnet and the longitudinal axis of the
patient. This is true for most superconducting
magnets but not for most permanent magnets.
In a CT image, shades of gray are assigned on In keeping with earlier discussions, a horizontal
the basis of Hounsfield units (HU), which relate orientation is assumed for the following
to x-ray attenuation values (Figure 14-2). White discussion.
is assigned to pixels with high attenuation values The Z-axis, which is parallel to B0, runs
(e.g., bone); black is assigned to pixels with low through the center of the magnet in the direction
attenuation values (e.g., air). of the long axis of the patient. The X- and Y-axes
In MRI, the gray scale is assigned on the basis are perpendicular to Z. X is horizontal and Y is
of the intensity of the MR signal emitted from a vertical (Figure 14-4).
given voxel (see Figure 14-2). That intensity, in B0 is intense and is energized throughout the
turn, is dependent on the T1, T2, and proton MRI examination. Vendors go to great lengths
density (PD) of tissue. White is assigned to pixels to ensure that this B0 field is not only intense, but
with high signal intensity; black is assigned to also uniform throughout the imaging volume.
those with low signal intensity. Subsequent chap­ Purposefully, the vendors also provide a means
ters fully discuss pixel character and therefore of disturbing the field homogeneity systemati­
image contrast. This chapter focuses on spatial cally and orderly.
localization. This disturbance of field homogeneity is
achieved through the use of paired gradient coils.
Gradient magnetic fields provide spatial localiza- These paired gradient coils are the key features
tion of the MR signal (RFs). distinguishing an MRI system from a nuclear
magnetic resonance (NMR) spectrometer (see
The intensity of the received MR signal (RFs) Chapter 11).
as it is apportioned among the pixels of the The gradient coils produce relatively weak
image determines its spatial location. A single magnetic fields superimposed on the B0 field.
signal is emitted from the entire slice after excita­ These gradient coils are intermittently pulsed for
tion by a slice-selective pulse, (RFt, BSS). milliseconds each time. Each pair of coils pro­
Even though each voxel has its own net mag­ duces a small gradient magnetic field along one
netization that produces an MR signal, the signal of the axes.
received by the imaging system (RFs) combines These small gradient magnetic fields are either
the signals from each voxel (Figure 14-3). This parallel or perpendicular to B0. Because the gra­
differs from radiography, which is basically a dient magnetic field varies in intensity along the
projection shadowgram. direction of the axis of the gradient coils, the
In most MRI systems, the main magnetic field combined magnetic field is stronger at one end
(B0) is horizontal and aligned with the bore of of an axis than at the other.
196 PART IV  Image Formation

is required to match the resonance frequency of

those protons. The combination of the magnetic
field inhomogeneity produced by the gradient
coils and the excitation by a specific RF fre­
Y quency permits slice selection.
During relaxation, the frequencies of the
signals from the protons also depend on the
magnetic field strength at the location of each
Z excited proton. When the RFt is turned off and
the excited protons relax back to equilibrium,
the free induction decay (FID) at the strong end
of a gradient magnetic field has higher frequen­
X cies than the FID at the weak end of the field.
A Consequently, the presence of a gradient mag­
netic field during signal reception helps localize
the protons within the slice that was selected
during excitation.

Z
SLICE SELECTION
If the Z gradient coils are on, the magnetic field
intensity varies linearly from one end of the
Y patient to the other (Figure 14-6). At one end of
the magnet, the magnetic field is stronger than at
the other end. The proton precessional frequency
B X for a 1 T magnet is 42 MHz. A proton at the
FIGURE 14-4  A, Axis orientation in the gantry. B, Axis strong end may precess at 43 MHz, whereas one
orientation for vector diagrams. at the weak end may precess at 41 MHz.
An RF pulse of exactly 42 MHz excites all
of the protons in the plane perpendicular to the
Z-axis at the 42-MHz position along the gradi­
GRADIENT COIL FUNCTION
ent magnetic field. At all other points along the
The purpose of the gradient magnetic field is gradient magnetic field, the protons remain unaf­
twofold: slice selection and pixel localization fected by the 42 MHz RF pulse because they do
within the slice. The gradient coils identify which not resonate at that frequency. They are insensi­
part of the signal belongs in each pixel. tive to it. In this way, the Z gradient performs
The Larmor equation (ƒ = γB) states that slice selection (GSS), and a single transverse slice
hydrogen nuclei in a magnetic field precess at a is selectively excited.
frequency (ƒ) that depends on the strength of the If a gradient along the X-axis were used
magnetic field (B). The constant γ is characteristic instead of the Z gradient, the slice selected would
of a given nuclear species. For hydrogen, the be in a sagittal plane. Similarly, a Y gradient
constant equals 42 MHz/T. If the magnetic field would select a coronal plane. In fact, with the X,
is not homogeneous, protons in a slightly stron­ Y, and Z gradients turned on together, any plane
ger field precess faster than those in a weaker in the body may be chosen (see Figure 14-6). All
magnetic field (Figure 14-5). subsequent discussions assume transverse slice
For protons to be excited in an area with a selection; however, the principles apply to slices
stronger magnetic field, a higher frequency, RFt, in any orientation.
 CHAPTER 14  The Musical Score 197

ω1 = γB1

ω2 = γB2

RF Slice
Gradient +
(excitation) selection

Emitted signal Pixel localization

Gradient +
(relaxation) within slice
FIGURE 14-5  Gradient magnetic fields cause slight differences in proton resonant frequencies, which can be used for
selective excitation of a slice and for localization of protons in that slice.

Y
B0
Z
X

No gradient

Z gradient on X gradient on Y gradient on

X + Y + Z gradients on
FIGURE 14-6  Selective excitation along the Z-axis results in a transverse plane. Excitation along the X- and Y-axes results
in parasagittal and paracoronal planes, respectively. When the three gradients are energized at the same time, any oblique
plane can be selected.
198 PART IV  Image Formation

The steepness, or slope, of the gradient mag­ 42  0.5 MHz

netic field and the range of frequencies, or band­
width, of the RF pulse determine the thickness
of the selected slice. The steeper the gradient
magnetic field, the thinner the slice. The nar­
rower the bandwidth of the RF pulse, the thinner
the slice.
The purity of an RF pulse is identified by its
quality value (Q). The Q of a pulse is the center BZ
frequency divided by the bandwidth. As band­
width becomes more narrow, the Q of an RF
pulse becomes higher.
Bandwidth is the range of frequencies con­
tained within the RF pulse (RFt). This is different 42  0.1 MHz
from the receiver bandwidth, which was dis­
cussed in Chapter 13. Ideally we would like
to make the RFt arbitrarily small. However,
achieving narrow bandwidth comes at a cost.
The inverse relationship between time and fre­
quency space is in play for RF pulse production
as well as for RF signal sampling. This means
that in order to make a slice narrow in thickness BZ
the selective RF pulse used must play out for a
long time.
FIGURE 14-7  In the presence of a fixed Z gradient mag-
netic field, reducing the radiofrequency (RF) bandwidth
The bandwidth of an RF pulse (RFt) becomes increases the RF Q value, which results in thinner image
narrower as the duration of the RF pulse is slices.
increased.

Long RF pulses cause problems for several

reasons. First, the longer the RF pulse then Answer: The frequency bandwidth is RFt =
the longer the minimum TE that will be possible 20 kHz * 2 ms/4 ms = 10 kHz and the slice
for a given pulse sequence. Second, if the thickness is 6 mm * 2 ms/4 ms = 3 mm.
RF pulse gets very long then T2 relaxation
can cause the signal to decay before the pulse In the presence of a fixed Z gradient magnetic
has even ended. We can see that a trade-off field, slice thickness can be selected by changing
must be made so that we have reasonably narrow the bandwidth of the excitation RF pulse (Figure
slice thickness and reasonably short TE values. 14-7). Alternatively, slice thickness can be selected
This explains why the minimum slice thickness by changing the intensity of the gradient mag­
available on MRI scanners is rarely less than netic field in the presence of a constant Q RF
3 mm. excitation pulse (Figure 14-8).
Gradient magnetic field intensity ranges from
Question: An RF pulse with RFt = 20 kHz is approximately 10 mT/m to 40 mT/m. The differ­
2 ms long and produces a 6 mm slice. What ence is due to the power supply that drives
is the RFt and slice thickness produced by that the direct current through the gradient coils. The
same RF pulse if its duration is increased to steeper the applied gradient magnetic field, the
4 ms? thinner the imaged slice.
 CHAPTER 14  The Musical Score 199

42  0.3 MHz

6 mT/m

BZ

42  0.3 MHz

12 mT/m

BZ

FIGURE 14-8  For a given radiofrequency pulse, slice thickness is thinner with stronger gradient magnetic fields.

Usually the frequency width of the RF excita­ in the spatial domain or pixel position (see
tion pulse is fixed for a range of slice thicknesses Chapter 13).
(e.g., 1 to 10 mm), and the strength of the slice MRI does not measure spatial location directly
selection gradient magnetic field, BSS, is adjusted like most medical imaging modalities. Instead, it
to fine-tune the slice thickness within the range measures spatial frequency content directly.
allowed by the particular RF excitation pulse. Spatial location must be determined by analyzing
the spatial frequencies during the image recon­
struction process.
PIXEL LOCATION WITHIN A SLICE
Perhaps the most difficult concept to grasp in Pixel location is determined by the inverse Fourier
MRI is how to determine the location of each transform of k-space.
pixel within a slice. This is because the MR data
are acquired in the spatial frequency domain, The spatial frequencies of the FIDs from the
k-space, whereas radiologists visualize the patient two highlighted voxels shown in Figure 14-9
200 PART IV  Image Formation

have the same frequency but different ampli­ The frequency-encoding gradient magnetic field
tudes. The frequencies are the same because the or read gradient (BR) is energized during signal
voxels are in a uniform magnetic field. A gradient reception.
magnetic field is not present. The higher-
amplitude signal occurs because of either a higher
PD or a different relaxation time (T1 or T2). This discussion concentrates on the two-
When a gradient magnetic field is energized dimensional Fourier transform (2DFT) imaging
during relaxation, pixels can be localized by their technique, which is the predominantly used
characteristic frequency along the X-axis (Figure method in clinical MRI. In 2DFT imaging,
14-10). In the same way, columns are localized k-space is sampled more quickly, much as a tele­
within a slice by energizing a frequency-encoding vision picture is displayed line by line on the
gradient (Figure 14-11). television tube.
Similarly, the phase-encoding gradient mag­
netic field (BΦ) localizes pixels into rows (Figure
14-12). The phase-encoding gradient controls
which horizontal line of k-space is to be sampled;
the frequency-encoding gradient accomplishes
the actual sampling across the line of data.

No gradient FID
THE EFFECT OF A GRADIENT
FIGURE 14-9  Without a gradient magnetic field present,
signal frequency from all voxels is the same and spatial
ON PRECESSION
localization of the signal is impossible. In the discussion regarding how protons precess
in the presence of a magnetic field, the effect of
the static magnetic field (B0) is ignored. This
viewpoint is called the rotating frame of refer-
ence (see Chapter 3).
In the rotating frame of reference, all protons
are stationary unless in a magnetic field that is
FID
stronger or weaker than B0, such as a gradient
magnetic field. If protons are in a field stronger
than B0, they precess at a higher frequency. If the
X gradient on
protons are in a field weaker than B0, they precess
FIGURE 14-10  With the X gradient magnetic field ener-
gized during relaxation while receiving a free induction with lower frequency (Figure 14-13).
decay (FID), frequency differences make pixel location by Consider the evolution of the net magnetiza­
column possible. tion vectors of the voxels along the direction of

Computer
Received MRI Fourier Columns
signal transformed localized
FIGURE 14-11  The free induction decay received while the X gradient magnetic field is energized is Fourier transformed
to provide information about the column of origin of the signal contribution.
 CHAPTER 14  The Musical Score 201

FIGURE 14-12  As the Y gradient magnetic field increases in intensity, the phase shift between adjacent rows also
increases.

along the gradient has a unique precessional

frequency.

BΦ Phase Encoding
What happens to the net magnetization vectors
of each voxel if a gradient magnetic field is
B0
applied for only a short time—a pulse? Before
the application of the gradient, all magnetization
Laboratory frame vectors precess in phase at the same frequency
(Figure 14-14).
During the application of the pulsed gradient
magnetic field, the net magnetization vector of
each voxel precesses with a frequency determined
by its position along the gradient. When the
pulsed gradient is turned off, all magnetization
vectors return to precessing at the same fre­
quency. The effect of the pulsed gradient is still
present in the form of a difference in phase for
BΦ each net magnetization vector.

The phase-encoding gradient magnetic field, BΦ,

is energized as a pulse before signal reception.
Rotating frame
FIGURE 14-13  Proton spins precess with increasing fre- The phase shift in the voxels changes as a
quency in the presence of a gradient magnetic field in the result of their position. This is a spatial frequency.
laboratory frame representation. Such spins precess with The originators of the 2DFT method, William
either positive or negative frequency from the average in Edelstein and James Hutchinson, called it the
the rotating frame representation. spin warp method because the phases of the net
magnetization vectors twist along the direction
the gradient magnetic field. The voxels on of the gradient, just as a warped piece of lumber
the positive end of the gradient precess counter­ twists along its length.
clockwise; voxels on the negative end of the gra­ An object can be viewed as a sum of objects,
dient precess clockwise. Voxels further from the each of which has a single spatial frequency
center of the gradient precess faster. Each point (see Chapters 12 and 13). When one of these
202 PART IV  Image Formation

voxel reinforce each other, resulting in a signal

of significant amplitude.
Ordinarily, the only spatial frequency observed
is the zero-frequency component because it
Same frequency
does not have an inherent phase twist. When
phase encoding is applied, the spatial frequency
with no net twist and therefore no signal cancel­
No gradient lation is controlled by the phase-encoding gradi­
ent pulse.

The duration of the phase-encoding gradient

pulse is fixed, and the amplitude is varied.

A stronger gradient magnetic field pulse

Same frequency makes all of the spins precess proportionately
faster. In a fixed amount of time, more twist per
BΦ
unit distance along the direction of the gradient
is produced.
For a particular spatial frequency with the
phase-encoding gradient pulse to be measured,
Pulsed phase-
encoding gradient
its amplitude is chosen so that the desired spatial
frequency is exactly untwisted. In actual imaging
FIGURE 14-14  Each spin system precesses at the same
frequency but with different phase after a pulsed phase- pulse sequences, the phase-encoding gradient
encoding gradient. pulse is stepped through the spatial frequencies
so that they are measured individually. As
a result, the resolution in the phase-encoded
direction of the image directly relates to the
frequencies is phase encoded, the twist or warp number of measurements made with different
from the phase encoding is added to the twist phase-encoding values resulting from different
inherent in that spatial frequency. intensities of the phase-encoding gradient mag­
If the original twist equals and opposes the netic field.
phase-encoding twist, then the resulting object is
straight as if the warped board was ironed flat.
Frequency Encoding
If the original twist and the twist from phase
encoding are not equal and opposite, the result­ Phase encoding takes care of one direction within
ing object is twisted even more. the slice, but the slice is two-dimensional. As a
Remember that the MR signal is a bulk signal result, it is necessary to measure the spatial fre­
from the excited spins in the sensitive volume of quencies in the direction perpendicular to the
the receiver coil (see Chapter 5). The signal from phase-encoded direction. This is done by apply­
each voxel adds vectorially to produce a single ing a gradient magnetic field in the direction
signal for the entire excited slice. within the slice that is perpendicular to the direc­
This signal has profound implications for tion of the phase-encoding gradient.
imaging. If an object has a phase twist in it, the This second gradient is left on while acquiring
phase shifts of the voxels are different. When the the MR signal (RFs) and is called the read gradi-
phase shifts are added vectorially, the result is ent (BR). For the following discussion, ignore any
essentially nil. However, if a twist is not present, previous phase encoding and consider only the
then the vector sum of the signals from each effect of frequency encoding.
 CHAPTER 14  The Musical Score 203

When a gradient magnetic field is first applied,

the voxels have net magnetization vectors that
are all aligned and also have spatial frequencies
with the normal inherent twist per unit distance.
As the frequency-encoding gradient is applied,
the net magnetization vectors begin to precess
faster or slower proportional to their locations
along that axis. In other words, the phases of the
voxels begin to twist.
Throughout the duration of the frequency-
encoding gradient pulse, the amount of twist
increases linearly with time. Recall from phase
encoding that applying a phase twist causes a
different spatial frequency to be untwisted and
measurable.
In the case of the frequency-encoding gradi­
ent, the twist is applied continuously, thereby
changing the measurable spatial frequency con­
stantly. This means that if the signal throughout
the duration of the frequency-encoding gradient
pulse is measured, a sampling of all spatial fre­
quencies along the frequency-encoded direction
results.
FIGURE 14-15  Preparing a potato for potato chips,
Note that only half of the spatial frequencies french fries, and diced potatoes is similar to the application
are observed because the sweep starts at zero of the three gradient magnetic fields.
spatial frequency. For the measurement to be
started at the edge of the spatial frequency plane
rather than in the middle, a pretwist is applied sweep of horizontal spatial frequencies associ­
to the object before the actual frequency-encoded ated with each row is measured with different
measurement is made. vertical spatial frequencies.
If a gradient pulse of half the duration and
opposite amplitude is applied, the phase shifts of The application of the three gradient magnetic
the object in the frequency-encoded direction are fields—BSS for slice selection, BΦ for phase
initialized to the maximum negative spatial fre­ encoding, and BR for frequency encoding (the
quency, so that the sweep covers both halves of read gradient)—provides the spatial localization
the spatial frequency information. necessary to identify each voxel.
Phase encoding selects a particular spatial fre­
quency in one direction within the selected This process is similar to that used to prepare
slice—the vertical direction. Frequency encoding a potato (Figure 14-15). First the Z gradient
then scans all of the spatial frequencies perpen­ is used to select a slice, like slicing a potato to
dicular to the phase-encoded direction, horizon­ make potato chips (BSS). Next, the phase-encoding
tally. As a result, all of the horizontal spatial gradient is energized momentarily, allowing the
frequencies at a particular vertical spatial fre­ columns to be identified within the slice (BΦ). If
quency have been sampled. the potato slice is thick enough, these columns
For the entire k-space to be measured, this would become french fries. Finally, the frequency-
process is repeated with different amplitudes encoding gradient is energized during signal
of the phase-encoding gradient each time. The acquisition, which identifies the rows in each
204 PART IV  Image Formation

Partial saturation
90
Bss RFt
FIGURE 14-16  A shaded trapezoid is used to symbolize Bss
the gradient magnetic fields.
RFs

BΦ
column (BR). For the potato analogy, the result
is diced potatoes. BR
FIGURE 14-17  The first signal acquisition of a partial
saturation pulse sequence.
PULSE SEQUENCE DIAGRAMS
This chapter started by suggesting the analogy
between the musical score used to conduct an pulsed at the beginning of the FID. This is shown
orchestra and the pulse sequence diagrams of the on the fourth line of the musical score as BΦ.
MRI system. Then, for a continuation of the During the signal acquisition, the phase-
analogy, the piece of music itself was explained. encoding gradient magnetic field (BΦ) is turned
The last section of this chapter discusses some of off, and the frequency-encoding read gradient,
the different types of pulse sequences (i.e., differ­ (BR), is turned on. The result is the five-line pulse
ent musical forms). sequence diagram shown in Figure 14-17.
One signal acquisition samples only one row
of k-space. Typically, 256 such acquisitions are
Partial Saturation needed to make an image. Each time a new signal
One simple pulse sequence is called partial satu­ is acquired, the strength of the phase-encoding
ration. In a diagram of this pulse sequence, the gradient changes (Figure 14-18).
top line indicates when to turn on the RF trans­ Because the only change from one signal
mitter, specifying a 90° RF exciting pulse (RFt). acquisition to the next is the intensity of the
The symbol for this RF pulse, called a sinc pulse, phase-encoding gradient magnetic field, the
is shown in Figure 13-9. The numerical notation phase-encoding gradient is represented as a
above the sinc symbol indicates the magnetiza­ stepped-pulse envelope (Figure 14-19). This
tion flip angle. implies a repetitive application of this basic pulse
For a single transverse slice excitation the Z sequence with variation only in the amplitude of
gradient magnetic field, BSS, must occur at the the phase-encoding gradient pulse.
same time as the RFt, and the frequency of
the RFt must be specified. This is indicated by the The RFt pulse sequence for partial saturation
trapezoid shape in Figure 14-16. imaging is 90°…//….90°..//…. 90°….//…. →.
Although the gradient magnetic fields are
thought to be turned on and off instantaneously, Finally, the time interval between each 90°
they are not. Some finite time is required to rise pulse, called the repetition time (TR), must be
to maximum intensity (i.e., slew rate) and then specified. The maximum duration of TR can be
return to zero again. These rise and fall times for any length of time, and its value typically
result in the trapezoid symbol, rather than a rect­ ranges up 10,000 ms; this time interval substan­
angular shape. tially influences the character of the pixels and
The FID signal is immediately observed after the image contrast. The TR and the number of
turning off the RFt. This is the third line of the phase-encoding steps have the dominant influ­
musical score, and it is labeled RFs for the MR ence on the overall time required to acquire
signal. The phase-encoding gradient is briefly the image.
 CHAPTER 14  The Musical Score 205

Partial saturation
TR
90° 90° 90°
RFt

BSS

RFs

BΦ

BR

Acquisition 1 Acquisition 2 Acquisition 3

FIGURE 14-18  Multiple partial saturation projections are obtained by varying the strength of the phase-encoding gradi-
ent magnetic field.

90° (2500 ms) × 1 × 256 = 640 s or almost

RFt 11 minutes

BSS If the resulting image appears too noisy, the

whole sequence can be repeated, and both sets
of data can be combined to make one image. This
RFs
procedure is called signal averaging.
Image quality can be improved if several
BΦ acquisitions are combined at the expense of
increasing the overall imaging time. The number
BR of signals averaged is symbolized by ACQ, for
FIGURE 14-19  The change in intensity of the phase- the number of signals acquired, or NEX, for the
encoding gradient magnetic field is represented symboli- number of excitations, or NSA, for the number
cally in a single pulse sequence. of signals acquired. If the number of signals
acquired in the example above were doubled,
the signal-to-noise ratio (SNR) would improve
Imaging Time
1
by 41% ( = 1.41), but the examination time
Time = TR × # BΦ × ACQ 2
where TR = repetition time, # BΦ = number of would double to nearly 22 minutes.
phase-encoding steps, and ACQ = number of
signals acquired per phase-encoding step Signal-to-Noise Ratio

Question: How long does it take to make an 1

S = 1, N = 1 SNR = =1
image, assuming a TR of 2500 ms, 1 signal 1
acquisition per phase-encoded step, and a 2
S = 2, N = 2 SNR = = 1.41
256 × 256 matrix? 2
Answer: Image time per slice = (TR) × (number 4
of signals acquired each phase-encoding step) S = 4, N = 4 SNR = =2
4
× (number of phase-encoding steps) =
206 PART IV  Image Formation

Partial saturation multislice

TR = 1000 ms

90° 90°
RFt (f1)

BSS

RFs

90° 90°
RFt (f2)

BSS

RFs

90° 90°
RFt (f3)

BSS

RFs
FIGURE 14-20  Multislice techniques used to reduce total imaging time.

Question: A given pulse sequence is repeated between signal acquisitions during which nothing
four times to improve contrast resolution by happens. That dead time during TR can be used
increasing SNR. How much is SNR improved? to acquire data from other slices. This is similar
Answer: At the expense of quadrupling examina­ to a musical round, that is, a song sung by several
tion time, signal(s) is also quadrupled to 4S. people who each start and end the same song at
Noise (N) is only doubled 4N = 2N. different times.
4S If the TR is 1000 ms and the FID for the first
Therefore SNR = = 2.
2N projection is gone after a few milliseconds, there
is a long waiting period while the excited protons
For a 10-slice head image, 11 minutes per slice relax to equilibrium. Interleaved multislice acqui­
would require nearly 2 hours of imaging time. sition takes advantage of that time by energizing
Although TR might be shortened to reduce the the RF transmitter with a different frequency to
imaging time, that would alter the contrast in the excite a different tissue slice.
image. The number of phase-encoding steps This RF pulse does not disturb the first
could be reduced, but that would reduce the slice because the frequency is inappropriate for
spatial resolution. the first slice. Because each slice excitation
It is possible to interleave the acquisition and data measurement only require a few milli­
of data from many different slices so that the seconds, the procedure can be repeated many
overall acquisition time equals that of a single times before the TR of the first slice is finished
slice, a technique called multislice imaging. Note and must be restimulated to acquire the next
that Figure 14-20 shows a significant time signal.
 CHAPTER 14  The Musical Score 207

Inversion recovery

TR

TI
180° 180° 180°
90° 90° 90°
RFt

BSS

RFs

BΦ

BR

Acquisition 1 Acquisition 2 Acquisition 3

FIGURE 14-21  The inversion recovery pulse sequence requires a 180° RF pulse followed by a 90° RF pulse.

Unfortunately, it is impossible to confine the (Figure 14-21). This pulse sequence helps to
RF excitation to a defined slice. The edges of better understand MRI contrast mechanisms.
the slice usually receive less RF energy than the Inversion recovery is also used in combination
middle of the slice. As a result, protons outside with fast imaging techniques, such as STIR and
of the effective slice thickness receive a slight FLAIR (see Chapter 16), which exclude signals
exposure to RF excitation. from fat and CSF, respectively.
The significance of this broadened excitation This pulse sequence, as its name implies,
is that when the time comes to excite the adjacent inverts the net magnetization vector by applying
slice, some of the protons have already been a 180° RF pulse. This inverting pulse takes the
excited and are undergoing relaxation. This leads net magnetization vector, which lies parallel to
to a degradation of the image contrast. the Z-axis at equilibrium, and rotates the vector
This problem of slice overlap (also called slice so that it points in the opposite (i.e., negative)
cross-talk) can be avoided with a nonimaged gap direction along the Z-axis. The spin-lattice relax­
specified between slices. If contiguous slices are ation time (T1) can be determined by timing how
needed, the multislice acquisition is specifically long it takes the inverted net magnetization
designed for maximum time between excitation vector to relax to equilibrium (see Chapter 7).
of adjacent slices, given the constraints of the The time interval between the 180° RF pulse
prescribed TR, a strategy known as interleaved and the 90° RF pulse is the TI. As with other
slice acquisition. time intervals, the duration of the TI interval
influences image contrast considerably.
Inversion Recovery The RF pulse sequence for inversion recovery
The inversion recovery pulse sequence has inher­ imaging is 180°…90°…//…180°… 90°…//
ently long imaging time because TR is long …180°…90°…//…→.
208 PART IV  Image Formation

Inversion recovery multislice

TR = 1000 ms

TI = 400 ms
180° 180°
90° 90°
RFt (f1)

RFs

180° 180°
90°
RFt (f2)

RFs

FIGURE 14-22  Multislice techniques can be used with the inversion recovery pulse sequence; however, the number of
slices is limited.

As with the partial saturation pulse sequence, echo sequence and differs from the standard spin
the Z gradient magnetic field (BSS) must be on echo sequence only in that the echo time (TE) is
during the 180° and 90° RF pulses for proper as short as possible. Similarly, the inversion
slice selection. The BΦ is pulsed and the BR is on recovery pulse sequences, used for MRI, are actu­
while receiving the FID to provide spatial local­ ally inversion recovery spin echo sequences with
ization within the slice. Although any time inter­ short TE.
val shorter than TR can be used for TI, typical The use of the FID as the MR signal in gradi­
durations range from 100 to 2000 ms. ent echo (GRE) imaging will be revisited in
Multislice techniques can also be used with Chapter 18.
the inversion recovery pulse sequence (Figure
14-22). However, the long TI interval severely
Spin Echo
restricts the number of slices that can be obtained
during a TR interval. For example, if the TR is The spin echo pulse sequence is the most com­
2500 ms and the TI is 400 ms, only five slices monly used MRI pulse sequence (Figure 14-23).
can be interleaved. In spin echo imaging, the proton spins are
For instructional simplification, the pulse pretwisted as the first step of frequency encoding
sequences were introduced with the assumption in the 2DFT method. For a spin echo to be
that FIDs are collected to make MR images. obtained, first, a 90° excitation RF pulse is used
However, the switching on and off of the gradi­ to flip the magnetization into the XY plane. This
ent magnetic fields and other practical timing is followed by a 180° refocusing RF pulse that
considerations make it difficult to acquire a true flips the vectors about an axis in the XY plane.
FID, so an MR signal called a spin echo is After time, an echo forms.
acquired instead.
The spin echo is the most common kind of The RF pulse sequence for spin echo imaging is
echo signal. In fact, a partial saturation pulse 90°…180°…//…90°….180°…//…90°…180
sequence is actually a partial saturation spin °….//…→.
 CHAPTER 14  The Musical Score 209

Spin echo

TR
TE
1/2 TE
180 180
90 90
RFt

BSS

RFs

Bφ

BR

Acquisition 1 Acquisition 2

FIGURE 14-23  The spin echo pulse sequence requires a 90° RF excitation pulse followed by a 180° refocusing RF pulse.

For the spin echo pulse sequence and most must also be specified because it too is an impor­
other pulse sequences, spin echoes are the MR tant determinant of image contrast. In clinical
signals used to make an image. As before, the MRI, TEs can range from approximately 2 to
slice selection gradient magnetic field (BSS) 120 ms.
remains on during the 90° excitation RF pulse
for slice selection. When the 90° pulse is turned Question: What is the imaging time for a spin
off, an FID is formed but ignored. echo pulse sequence with TR = 2000 ms,
After 1/2 TE, a 180° refocusing RF pulse is TE = 80 ms, ACQ = 2, and a 256 frequency-
applied. The slice selection gradient is turned on encoding samples × 192 phase-encoding
again during the 180° pulse so that only those steps?
protons in the slice of interest are affected. At TE Answer: Image time = (TR) × (number of signal
the spin echo forms. acquisitions) × (number of phase-encoding
The spin echo forms at TE after the 90° RF steps)
pulse. Because the spin echoes are the MR signals = 2000 ms × 2 × 192
acquired to make the image, the read gradient = 768 s
(BR) must be on during the spin echo to encode = 12 min 48 s
spatial localization information.
For multislice imaging, the spin echo must be
The time interval between the center of the obtained before exciting the next slice (Figure
90° RF pulse and the center of the spin echo is 14-24). If TR equals 1000 ms and TE equals
the TE. 30 ms, excitation of the next slice can be imple­
mented after about 50 ms; for this situation, a
Once again, the interval from one 90° RF maximum of 20 slices can be excited before
pulse to the next, the TR, must be specified returning to the first slice for the second phase-
and is important for pixel character. The TE encoding step.
210 PART IV  Image Formation

Spin echo multislice

TR = 1000 ms

TE = 30 ms
90 180 90 180
RFt (f1)

RFs

TE = 30 ms
90 180
RFt (f2)

RFs

FIGURE 14-24  The radiofrequency pulse sequences necessary for multislice spin echo imaging.

It is often clinically useful to make images excited within the 2500-ms TR interval. In this
with different spin echo times, a technique called way, it is possible to generate 50 images in the
multi-echo imaging. A pulse sequence with TR time previously required to make one.
equal to 2500 ms, TE equal to 30 ms, 256 pro­ This common imaging technique is called mul­
jections, and 1 signal average takes 10 min, 40 s. tislice, multi-echo spin echo imaging. The musical
It may be useful to have another image made score* is complete (Figure 14-26).
with the same parameters except for making TE
equal to 60 ms, which would take an additional
CHALLENGE QUESTIONS
10 min, 40 s.
However, both images can be obtained in one 1. How many lines of information are there in
pulse sequence by taking advantage of the echo an MRI pulse sequence diagram and what
formed by a second 180° RF refocusing pulse are they?
(Figure 14-25). At 15 ms after the 90° RF excita­ 2. How long does it take to acquire a gradient
tion pulse, a 180° RF refocusing pulse is applied echo (GRE) image having the following
to produce a spin echo with a TE of 30 ms; characteristics: TR = 20 ms, TE = 6 ms,
after an additional 15 ms, another 180° RF refo­ number of acquisitions = 2, and the matrix
cusing pulse produces a second spin echo with a size is 512 × 512?
TE of 60 ms. 3. In radiography, kVp controls image con­
The 256 × 30 ms spin echoes are collected to trast. What principally controls contrast in
make one image; the 256 × 60 ms spin echoes MRI?
are collected separately to make another image. 4. Diagram the RF pulse sequence for an inver­
This results in two separate images with different sion recovery image and indicate the timing
image contrast because of the different TEs. for the appearance of the MR signal.
Because both echoes are collected within
60 ms, excitation of another slice can begin at *Susan Weathers, MD, is the author of this felicitous
approximately 100 ms; therefore 25 slices can be metaphor.
 CHAPTER 14  The Musical Score 211

Spin echo
multiecho

TR = 1000 ms

RFt

RFs
TE1 = 30 ms

TE2 = 60 ms

Acquisition 1

FIGURE 14-25  Multi-echo imaging allows the acquisition of distinctively different images formed at different echo times.

Multislice
multiecho
spin echo
RFt
BSS
RFs Slice 1
Bφ
BR
RFt
BSS
RFs Slice 2
Bφ
BR
RFt
BSS
RFs Slice 3
Bφ
BR
RFt
BSS
RFs Slice 4
Bφ
BR

RFt
BSS
RFs Slice N
Bφ
BR
FIGURE 14-26  Multislice, multi-echo spin echo imaging involves an amazing arrangement of radiofrequency and gradi-
ent magnetic field pulses.
212 PART IV  Image Formation

5. For 2DFT MRI, why are contiguous slices 9. For most superconducting MRI systems, the
not obtained as in CT? B0 field is horizontal. For most permanent
6. Three signals are acquired and summed to magnet MRI systems, the B0 field is vertical.
improve SNR. How much improvement will For all MRI vector diagrams, the Z-axis is
be realized? vertical. What is the relationship between B0
7. When is the slice selection gradient coil (GSS) and the Z-axis?
energized? 10. What is the difference between a gradient
8. How does the vector diagram appear coil (GXYZ) and a gradient magnetic field
when an ensemble of spins are partially (BSS, BΦ, BR)?
saturated?
 CHAPTER

15 
Magnetic Resonance
Images

OBJECTIVES
At the completion of this chapter, the student • Identify the three principal magnetic resonance
should be able to do the following: imaging (MRI) parameters.
• Describe the features of a visual image. • Name the primary and secondary MR signals.
• Define representational image and cite three • Relate the approximate spin echo pulse
examples. sequences necessary to obtain proton density–
• Discuss the pixel character of a magnetic weighted (PDW), T1-weighted (T1W), and
resonance (MR) image. T2-weighted (T2W) images.

OUTLINE
What is an Image? Image Evaluation Criteria Pure Magnetic Resonance
Visual Images Spatial Criteria Images
Optical Receptors Pixel Character Weighted Images
Pattern Recognition Magnetic Resonance Images Partial Saturation
Location and Character Magnetic Resonance Image Inversion Recovery
Representational Images Character Spin Echo
Radiographic Images Magnetic Resonance
Magnetic Resonance Images Imaging Parameters

KEY TERMS
Contrast perception Refocusing pulse Visual acuity
Null point Spatial resolution

the duck elicited by the sound of the oboe in

WHAT IS AN IMAGE?
Prokofiev’s Peter and the Wolf. An image can
In the most general sense an image is a mental also be abstract, such as the famous painting by
picture. It may be a visual image based on Tanner (Figure 15-1).
direct observation, such as the viewing of the Medical images generally fall into the category
Grand Canyon, or an imaginary image, such as of visual images; they attempt to represent real

213
214 PART IV  Image Formation

FIGURE 15-1  This famous painting by Tanner, entitled Stampede across the Pecos, is an example of an abstract image.
(Courtesy Raymond Tanner, Memphis, TN.)

objects accurately. A brief discussion of some to intensity and color. This initial information
basic concepts of visual images is helpful in about the source and character of light is then
explaining magnetic resonance (MR) images. relayed through complex visual pathways to the
occipital lobe, where the most complex process-
ing and analysis of these data are performed
Visual Images (Figure 15-2).
All visual images are initially detected by the In the occipital lobe, the myriad of discrete
eye and are the result of stimulation of receptor data concerning light impinging on the retina is
cells in the retina by electromagnetic radiation in reconstructed into the perceived image. Although
the visible region of the spectrum. The two types the final image is often considered to be a con-
of receptor cells are rods and cones. These recep- tinuum of information about light, it is initially
tor cells can be considered digital detectors that detected, processed, and conceived as a high-
are stimulated by the input of light photons. resolution digital image.
They respond to such stimulation with an output
of discrete electrical impulses called action
Optical Receptors
potentials.
Each retinal receptor is arranged by the When light arrives at the retina, it is detected by
optical structure of the eye to be sensitive to light the rods and the cones. Rods and cones are small
coming from a specific region of the visual field structures; there are more than 100,000 of them
and to characterize the nature of that light as per square millimeter of the retina.
 CHAPTER 15  Magnetic Resonance Images 215

CPU

Control unit

Memory

Arithmetic unit

FIGURE 15-2  Processing and analysis of signals from the visual receptors occur in the occipital lobe. This occurs much
like a computer.

The cones are concentrated on the center of to detect differences in brightness levels. This
the retina in an area called the fovea centralis. property of vision is termed contrast perception.
On the other hand, rods are most numerous on Furthermore, cones are sensitive to a wide range
the periphery of the retina. There are no rods at of wavelengths of light.
the fovea centralis.
The rods are very sensitive to light and are Cones perceive color; rods are essentially
used in dim-lighting situations. The threshold for color-blind.
rod vision is approximately 10−6 ml. However,
cones are less sensitive to light; their threshold is The cones of the eye are color receptors
only 5 × 10−3 ml, but they can respond to intense of three types, each stimulated by relatively
light levels, whereas rods cannot. Consequently, narrow bandwidths of light generally referred
cones are used primarily for daylight vision, to as red, blue, and green. They are sensitive
called photopic vision, and rods are used for to brightness (number of light photons), hue
night vision, called scotopic vision. (wavelength), and saturation (ratio of mono-
This aspect of visual physiology explains why chromatic to white light). The human eye can
dimly lit objects are more readily viewed if they detect approximately 2000 different color com-
are not looked at directly. Astronomers and radi- binations or hues compared with perhaps only
ologists are familiar with the fact that a dim 20 shades of gray.
object can be seen better if viewed peripherally, Therefore color images contain significantly
in which case rod vision dominates. more information than black and white images.
The ability of the rods to visualize small Imaging techniques such as magnetic resonance
objects is much worse than that of the cones. imaging (MRI) intrinsically contain more infor-
This ability to perceive fine detail is called visual mation and are more naturally adaptable to
acuity. Cones are also much more able than rods color imaging.
216 PART IV  Image Formation

In a standard clinical setting, when someone not, but at a distance, the dots appear to merge
views a 14 × 17 inch chest radiograph at 50 cm, into a continuum of space and color forming
the eye can spatially resolve 0.5 to 1 mm, shapes and patterns.
20 shades of gray, and 1000 different colors. The painting shown in Figure 15-4 is pat-
Optimal medical imaging should take maximum terned after the work of a famous pointillist,
advantage of this basic physiologic capability. Georges Seurat, and illustrates this concept.
Functional MRI (fMRI) is a first step in this When it is viewed at a distance, the painting is
direction. of a picnic in the Texas hill country and elicits
the psychological impression of such a real scene.
When it is viewed at close range, the painting
Pattern Recognition
loses its totality and becomes simply a matrix
The spatial map of light that is a visual image of dots.
does not have any intrinsic intellectual signifi- Such artists called these dots points, but these
cance. The intellectual value of such an image dots are perfectly analogous to what we call
primarily depends on the mental comparison of pixels. Each pixel has a unique location and
an image with the large library of images stored character. The location is the position of each
within the brain and the subsequent evaluation pixel in relationship to others. The character is
of the meaning of the image. represented by brightness, color, or both.
Most medical imaging is a process of pattern Most medical images differ from paintings in
recognition, which basically involves comparing what the character of the pixel means. When an
one image pattern with another. Evaluation for object is directly viewed, the pixel character is
similarities and differences with known patterns the actual brightness and wavelength of light
results in a “best fit” conclusion or “most likely” reflected to the eye. When a representational
diagnosis (Figure 15-3). image is viewed, the pixel character still primar-
This leads to the inescapable fact that a major ily reflects the light incident onto the eye, but this
factor in diagnostic efficiency is the observer’s initial character of light represents or stands for
development of a large image memory bank, something else. In a realistic painting, the pixel
which at least partially comes from viewing many character represents the visual light characteris-
images. Such a memory bank is currently imple- tics of actual objects.
mented and supplemented in CAD (computer-
assisted detection). Medical images are representational.
Regardless of the significance of an image,
visual images can be evaluated on the basis of
spatial resolution and contrast resolution. That Medical images represent something else
is, the perception and interpretation of an image and are not realistic. They do not attempt to
depend on the location of light photons and the copy a real, directly visualized object. For
differences in character of those photons. example, in the infrared images shown in Figure
15-5, the pixel character represents the radiant
heat of an object, a feature that cannot be seen
LOCATION AND CHARACTER directly.
An image consists of discrete spatial points, each Equally representational is the temperature
having different light characteristics. This concept map shown in Figure 15-6 in which the various
was appreciated and emphasized by a group pixels stand for the temperature of a region
of Postimpressionist painters called pointillists, rather than any geographic feature; such a rep-
who painted pictures made up of dots of differ- resentational image can be made whenever any
ent colors. When they are viewed at close range, parameter can be measured as a function of
the dots are obvious but what they represent is location.
 CHAPTER 15  Magnetic Resonance Images 217

A B

C D
FIGURE 15-3  Pattern recognition involves the mental comparison of images to fit the “best fit” pattern for diagnosis.
A, A clinically unknown case with high signal mass in the brain stem. B, Comparison pattern 1: low-signal brain stem
lesion without mass, indicating a hematoma. C, Comparison pattern 2: excentric, high-signal brain stem lesion without
mass, indicating multiple sclerosis. D, Comparison pattern 3: high-signal brain stem mass, indicating a brain stem glioma.

Representational Images
The character of a pixel can be made to stand
Medical images are representational because the for essentially any measurable quantity that can
character of the pixel does not attempt to reflect be spatially defined. Examples include electron
the actual visual light feature of the tissue, but density maps (radiographs), radioactive decay
rather, some other physical parameter that has maps (radionuclide images), and hydrogen con-
been measured and transformed into the image. centration maps (MR images).
218 PART IV  Image Formation

FIGURE 15-4  This scene of a picnic in the Texas hill country is patterned after the pointillist Seurat’s Sunday Afternoon
on the Island of La Grande Jatte. Essentially, it is a digital image. (Courtesy Frank Scalfano, Decatur, AL.)

Another obvious difference between most The basic difference between the radiographic
directly observed images and medical images is image and a direct visual image of the body is in
that medical images are of the interior of a body the devices that measure pixel character or the
rather than the surface. The primary clinician detectors. In the case of direct visualization, the
does the medical imaging of the surface of the detectors are the rods and cones of the retina. In
body by direct observation, whereas the radiolo- the case of a radiograph, the detectors are the
gist images the inside of the body. silver halide grains of the film emulsion that are
sensitive to short-wavelength electromagnetic
radiation.
Radiographic Images The number of x-rays detected determines
Imaging the inside of the body was a major pixel character. The amount of x-rays absorbed
problem in early medicine. Before Roentgen’s by the body is physically determined by the
discovery of x-rays in 1895, the only way medical x-ray attenuation coefficient, which is principally
images of the interior of the body were made was related to the density of electrons in different
by direct visualization. This obviously required tissues.
cutting into the subject and viewing the organ or
tissue of interest in the form of traditional medical
A radiograph is essentially an electron density
illustration (Figure 15-7).
map of the body.
With the discovery of x-rays, it became pos-
sible to make images of the interior of the body
in a relatively noninvasive fashion. Radiographic However, a radiograph does not relay any
images have features comparable to the represen- numerical information about the electron density
tational images discussed. A radiographic image (Figure 15-8). The informed observer knows
has pixels (groups of silver grains) with unique only that the blackness or whiteness on the image
location and concentration differences that is related to electron density. The whiter the
account for pixel character or brightness. image, the higher is the electron density.
 CHAPTER 15  Magnetic Resonance Images 219

B
FIGURE 15-5  These infrared images represent emitted radiant heat, which is something the optical receptors cannot
detect. A, A thermogram of a patient. B, An infrared photograph of the Texas Gulf Coast shows Houston at the top of
the photo and Galveston Bay at the upper right. (A, Courtesy Alphonso Zermeno, Austin, TX; B, Courtesy NASA.)
220 PART IV  Image Formation

FIGURE 15-6  This weather map is normally published in multicolor and shows regional temperatures.

FIGURE 15-7  Before radiography, the interior of the FIGURE 15-8  Radiographs are electron density maps of
body could be viewed directly only during surgery. tissue, but they do not provide numerical data. This
famous radiograph was taken by Roentgen and shows his
wife’s hand. (Courtesy Deutsches Roentgen-Museum.)

This gray scale information is very useful,

even if the physics of x-ray interaction and elec-
tron density are not considered. This concept is radiograph. The evaluation of bone remains one
used empirically to evaluate the internal struc- of the main uses of radiography.
ture of the anatomy. Radiographs differentiate among tissues and
Radiographic contrast resolution among therefore report gross anatomy. However, radio-
tissues is determined principally by differences in graphs can also relate pathologic anatomy by
electron density. For example, bone has a much showing abnormal gross anatomy, as reflected
higher electron density than soft tissue and is by deformities, distortions, absence, enlarge-
therefore highly contrasted to soft tissues on a ment, and reduction of organs and tissues.
 CHAPTER 15  Magnetic Resonance Images 221

function of some particular physical parameter.

Such medical images, then, are really three-
dimensional images. Two dimensions define pixel
location, and the third dimension, represented by
brightness or color, contains information about
the physical character of the tissue.

An MR image is essentially a map of the mag-

netic characteristics of nuclei within the atoms
in the body.

The physical character of the pixels in an MR

image is multidimensional. The numerical value
of an MRI pixel is principally determined by
the MRI parameters: proton density (PD), spin-
lattice relaxation time (T1), and spin-spin relax-
FIGURE 15-9  The lateral ventricles in this magnetic reso- ation time (T2).
nance image show such high contrast from surrounding Secondary determinants of MRI pixel values
brain tissue that further pixel characterization is largely are chemical shift, magnetic susceptibility, and
irrelevant. (Courtesy Pedro Diaz-Marchan, Houston, TX.)
motion. The weighting given to each of these
parameters can vary greatly depending on the
timing of the radiofrequency (RF) pulses and the
In addition to providing information about gradient magnetic fields.
normal and abnormal gross anatomy, radio-
graphs can provide some information about
IMAGE EVALUATION CRITERIA
physiology, such as the healing of fractures
reflected by calcium deposition in callus forma-
Spatial Criteria
tion or the abnormal deposition of calcium in the
basal ganglia in patients with hypercalcemia. The primary value of pixel location is to display
geometry or gross anatomy. The display of gross
anatomy requires not only appropriate pixel
Magnetic Resonance Images positioning but also adequate differences in
Unfortunately, most anatomical or physiologic pixel character to produce sufficient contrast
implications of radiographic images are indirect between tissues to allow their separation and
because there is no intrinsic biological signifi- recognition.
cance to electron density. The clinical significance However, most evaluation of gross anatomy is
of an x-ray image is based on the recognition of geographic and therefore relatively independent
patterns from accumulated experience of empiri- of the imaging technique. For instance, once the
cally evaluated radiographs. lateral ventricles have been displayed on an
Essentially, all forms of medical imaging use image (Figure 15-9), the character of the pixels
the substitution of some primary physical detec- becomes largely irrelevant.
tor system other than the human visual system.
The visual system becomes a secondary receptor The three principal spatial criteria are size, shape,
that interprets the representational images and position.
created by the primary receptor system.
All medical image detector systems create a The relevant factors for the evaluation of the
spatially defined map of the human body as a gross anatomy of the lateral ventricles or any
222 PART IV  Image Formation

A B C
FIGURE 15-10  These three spinal cord images demonstrate image evaluation according to object size. A, Normal.
B, Large. C, Small. (Courtesy Pedro Diaz-Marchan, Houston, TX.)

A B C
FIGURE 15-11  A, Normal quadrigeminal plate. B, An intrinsic deformity (glioma). C, An extrinsic deformity (hydrocepha-
lus). (Courtesy R. Nick Bryan, Philadelphia, PA.)

other structure are basically the geometric factors: extrinsically displaced. For example, the low
size, shape, and position. In the viewing of the position of the cerebellar tonsils in the Arnold-
image, the size, shape, and position of the lateral Chiari malformation is an intrinsic malposition,
ventricles are referenced to normal, which is usually indicating a congenital anomaly. On the
based on experience. other hand, the caudal herniation of the tonsils
Evaluation of the spatial aspects of an image from a posterior fossa mass is an extrinsic dis-
is relatively straightforward. The specific possi- placement, indicating the presence of a distant
bilities in terms of size are normal, large, and mass in addition to the malpositioned tonsil.
small (Figure 15-10).
The main classifications regarding the shape
Pixel Character
of a tissue are normal (Figure 15-11, A), intrinsi-
cally deformed, and extrinsically compressed. Unexpected pixel brightness indicates abnormal
For example, Figure 15-11, B, shows the beak tissue character. The significance of this is
deformity of the quadrigeminal plate in a patient entirely dependent on the imaging technique.
with the Arnold-Chiari malformation. This is an Pixels that are abnormally bright in the brain
intrinsic deformity, whereas the flattening of the on a computed tomography (CT) image suggest
quadrigeminal plate from a pineal tumor seen in either blood or calcium, whereas abnormally
Figure 15-11, C, is an extrinsic compression. bright pixels on T2-weighted MR images suggest
In terms of position, the main categories multiple sclerosis or tumor but not blood or
are normal, intrinsically malpositioned, and calcium.
 CHAPTER 15  Magnetic Resonance Images 223

and the presence of appropriate normal reference

tissue somewhere on the image. If inappropriate
pixel values are chosen for display, the lesion
may be excluded or windowed out of the image
(Figure 15-13). If an organ is diffusely involved
by disease, the abnormality may not be obvious
even if appropriately windowed because there is
no normal reference density (Figure 15-14).
In general, observers have more difficulty
detecting a minor overall change in pixel bright-
ness than detecting focal lesions. Despite the
lack of objectivity in the empirical evaluation
of images, it is still the primary interpretative
skill used.
The visual system is extremely well designed
for such empirical analysis. It is quick, and its
use in training is relatively inexpensive. Little or
FIGURE 15-12  Subjective evaluation of an image recog- no direct interaction is required between the
nizes brighter regions more easily than expected, as in the
observer and the instrument. More detailed,
case of cerebral glioma. (Courtesy R. Nick Bryan, Philadel-
phia, PA.) objective, and computer-assisted evaluation of
images (CAD) is now affecting and improving
clinical diagnosis.
Quantitative interpretation of medical images
As with spatial criteria of an image, evaluation requires knowledge of pixel values. This is an
of pixel character requires knowledge of what additional analytical step that can slow the inter-
is normal. There are basically two ways to pretive process and increase the complexity of
determine such normalcy. The first and most instrumentation.
commonly used is the subjective, empirical Images having pixel values set by various MRI
visual interpretation of the image. In this case, parameters such as T1, T2, etc. can be generated
the interpreter looks to see whether the pixels from more extensive sets of image data. Interpre-
simply appear to be brighter than expected tation of the quantitative data produced by these
(Figure 15-12). parametric maps also requires knowledge of the
actual normal range of numbers for the particu-
The three principal character properties are lar examining technique. This is not necessarily
normal, bright, and dark. required in the more subjective approach.
Obtaining quantitative data in a small portion
This technique works relatively well for expe- of an image by defining a region of interest (ROI)
rienced individuals looking at focal abnormali- is relatively easy and requires little interaction
ties. Here, the abnormal pixel brightness of the and instrument complexity. It is the most common
focal lesion is referenced to the presumed normal type of objective evaluation of an image.
pixel brightness of the adjacent normal tissue.
If the organ is symmetrical, such as the brain,
Magnetic Resonance Images
the normal reference tissue does not have to be
adjacent but can be in an equivalent position on The significance of abnormal pixel brightness
the opposite side of the body. in terms of tissue characteristics is directly a
This subjective approach depends on appropri- function of the imaging technique, which for
ate selection of pixel display values, windowing, MR images is the choice of RF pulses and
224 PART IV  Image Formation

A B
FIGURE 15-13  Improper windowing (A, width 849, center 469) may obscure an otherwise obvious lesion such as this
pinealoma (B, width 849, center 757). (Courtesy Michael Mawad, Houston, TX.)

A B
FIGURE 15-14  Even with proper windowing, diffuse disease may be missed because of a lack of adjacent normal tissue.
A, Fatty liver. B, Normal liver. (Courtesy Tom Hedrick, Houston, TX.)

gradient magnetic fields. Therefore, for each produce pixel values within specific ranges. In
imaging technique, the interpreter must know general, lesions that increase pixel values are dis-
which pathophysiologic processes increase pixel tinctly different from those that decrease pixel
brightness versus those lesions that diminish values.
pixel brightness. A number of practical things must be under-
At the subjective, empirical level, it often stood in the daily practice of MRI. Perhaps most
comes down to those lesions that appear very important are characteristics of an image and the
bright versus slightly bright versus slightly dark way they are influenced by MRI technique.
or very dark. With a quantitative approach, the The location of a pixel is usually well defined.
observer has to know the disease processes that The gradient magnetic fields are responsible for
 CHAPTER 15  Magnetic Resonance Images 225

A B
FIGURE 15-15  These brain images of multiple sclerosis show that the (A) magnetic resonance image is superior to the
(B) computed tomography image because of better low-contrast resolution.

pixel location. Without the gradients, there is no convey more information because three param-
spatial localization in MRI. eters (PD, T1, and T2) are involved. Neverthe-
The MRI technologist has little or no control less, at the present time, color MR images are
over pixel location. However, the MRI technolo- rarely produced except for fMRI.
gist is directly responsible for determining the
character or brightness of pixels.
MAGNETIC RESONANCE
The RF pulse sequence determines pixel character IMAGE CHARACTER
and therefore contrast rendition and contrast
A radiograph is a flat, gray image with low-
resolution.
contrast resolution and therefore has relatively
little detail. The tissue parameter that determines
Pixel character can be appreciated in two fash- brightness in this image is electron density, which
ions: black and white or color. In most medical varies by no more than 1% for most soft tissues.
imaging, black and white is used, but a color Therefore the inherent subject contrast is very
image can provide a great deal more information, low. Radiographic contrast is improved with
as in fMRI. grids, with the use of tomographic techniques,
In x-ray, gamma ray, and ultrasound images, and with the injection of radiographic contrast
the physical information that determines pixel material.
character can be adequately conveyed by a black The CT image shown in Figure 15-15, A, gives
and white format because only one parameter, even better contrast resolution than a radiograph
electron density, radioisotope concentration, or because the x-ray beam is highly collimated. This
reflectivity, respectively, is involved. collimation rejects scatter radiation, preventing
An MR image can be displayed in black and it from reaching the detectors. In the CT image
white, but a color rendition may potentially of the brain, the 0.5% difference between the
226 PART IV  Image Formation

µ
1000 PD 100 2000 2000
T1
T2 500 125
80
400 100
60

ms
ms
HU 0 300 75
40
200 50

20 100 25

−1000
Water Blood Heart Gray White Liver Bone Fat
Brain
FIGURE 15-16  This histogram shows that the relative differences in magnetic resonance imaging parameters for various
tissues are greater than the range of the x-ray linear attenuation coefficients of x-ray imaging as measured by Hounsfield
units (HU).

electron density of gray matter (GM) and white These MRI parameters can be discussed in
matter (WM) can just be detected. terms of an image without understanding PD,
The principal advantage of the MR image T1, or T2 at all. They are simply MRI character-
shown in Figure 15-15, B, over a CT image is istics of tissues. With proper technique, the MRI
contrast resolution. The anatomical location of technologist could potentially measure each of
the pixels is the same, but the character of the them as a function of location and create a para-
pixels is different. The improved contrast resolu- metric map or representational image.
tion is due to the intrinsic differences in the tissue Figure 15-16 is a histogram of x-ray linear
values of the MRI parameters. These parameters attenuation coefficients and the MRI parameters
differ by as much as 40%. PD, T1, and T2 for various tissues. The linear
attenuation coefficient of most soft tissues is
approximately the same.
Magnetic Resonance However, the MRI parameters of the same
Imaging Parameters tissues show a much greater variation. This vari-
The three principal MRI parameters are PD, T1, ation of MRI parameters among soft tissues
and T2. Secondary parameters such as chemical results in superior contrast resolution.
shift, paramagnetic materials, magnetic suscepti-
bility, and motion also influence pixel character.
Pure Magnetic Resonance Images
The MR imaging system detects these parameters
and converts them into pixel character. With color as an analogy, the MRI parameters
A complete MRI examination should include can be considered as being comparable to the
images of each of these parameters. However, three primary colors (blue, yellow, and red). All
such a complete MRI examination is extremely color images viewed consist of a mixture of these
time-consuming. With most MRI systems, chem- primary colors. Depending on the intensity of
ical shift imaging, paramagnetic contrast agents, each primary color in the mix, all colors of the
and motion imaging are reserved for specialized spectrum can be obtained.
exams. The relative contributions of the three A look at the screen of a color television con-
main tissue characteristics, PD, T1, and T2 are firms this. However, a magnifying glass may be
routinely detected in clinical MR images. necessary. The screen actually consists of only
 CHAPTER 15  Magnetic Resonance Images 227

FIGURE 15-17  A calculated pure proton density image. FIGURE 15-18  A calculated pure T1 image. (Courtesy
(Courtesy Richard Wendt III, Houston, TX.) Richard Wendt III, Houston, TX.)

three types of dots—blue, yellow, and red. Com- a bigger difference than the 0.5% difference in
binations of these dots produce the full-color the x-ray linear attenuation coefficients of the
image seen from a distance. Regions that appear same tissues as seen on CT images.
white have all dots glowing intensely. T1 Images.  Figure 15-18 shows a pure T1
image. The character of the pixels is influenced
only by the T1 of the tissue. The big difference
MR images representing solely PD, T1, or T2
between the T1 of GM and that of WM is
cannot be obtained.
obvious; this is a high-contrast image.
The T1 of GM is approximately 700 to
MR images consist of the influences of PD, 800 ms; on the other hand, the T1 of WM varies
T1, and T2. Pure MR images are analogous to from 500 to 600 ms, resulting in approximately
the primary colors because they represent only a a 30% difference in the T1 of these two tissues.
single MRI parameter. Unfortunately, in normal This difference produces a high-contrast image
practice, pure images of MRI parameters are not of the brain.
practical. Weighted images are routinely pro- T2 Images.  Figure 15-19 shows a pure image of
duced, and these are analogous to color images the third primary MRI parameter, T2. This looks
obtained by adding the primary colors in various significantly different from the other pure images
proportions. even though they are of the same slice from the
PD Images.  A pure PD image is shown in Figure same patient. There is little difference between
15-17. In this image, pixel brightness is a func- the T2 of GM and T2 of WM, and the result is
tion only of PD. The higher the PD is, the brighter very little contrast. The T2 of both GM and WM
the pixel. The GM has a higher PD than WM, is approximately 100 ms.
and therefore it is brighter. Routinely, these are the three primary MR
This image shows that a high-contrast MR images that should be obtained. Each of these
image can be made of the brain on the basis of three images is uniquely different from the others;
PD alone. There is about a 20% difference they may only occasionally look similar and
between the PD of GM and that of WM. That is happen to have similar contrast.
228 PART IV  Image Formation

inherent tissue contrast. Such weighted images

WEIGHTED IMAGES
have pixel intensity determined more strongly by
Clinically, RF pulse sequences are used to produce one of the three MRI parameters.
images that are a blend of the intrinsic PD, T1,
and T2 characteristics of tissue. When the three MR images are weighted by PD, T1, and T2.
primary colors are blended, hues that may
obscure original information are obtained.
Similarly, such an MR image may be practical The RF pulse sequences routinely used are
and aesthetically pleasing but may also obscure like color filters over a color photograph; they
can hide information. Sometimes that informa-
tion is not needed for diagnosis, but without it,
the diagnosis may not be certain. Partial satura-
tion, inversion recovery, spin echo, and gradient
echo RF pulse sequences result in weighted
images.
The following discussion deals only with
the RF pulse sequences because they determine
image contrast. It is assumed throughout that the
gradient magnetic fields have been pulsed prop-
erly to provide spatial localization and spatial
resolution.

Partial Saturation
Proton spins exist in four states, as shown in
FIGURE 15-19  A calculated pure T2 image. (Courtesy Figure 15-20. The patient is at equilibrium with
Richard Wendt III, Houston, TX.) the B0 field before the pulse sequence begins.

MZ = M0 MZ = 0

Equilibrium Saturated

MXY = 0 90RF MXY = M0

MZ = aM0

Partially
saturated
MXY = 0
180RF αRF MXY = bM0
Inverted

MZ = −M0
FIGURE 15-20  These vector diagrams represent the four spin states: equilibrium, saturation, inversion, and partial
saturation.
 CHAPTER 15  Magnetic Resonance Images 229

TR
90 90 90
RFt

RFs

MZ

FIGURE 15-21  A partial saturation radiofrequency (RF) pulse sequence is the simplest imaging sequence. Free induction
decays after the first repetition time (TR) have lower intensity because the proton spins remain partially saturated; that
is, all of them have not relaxed to equilibrium.

After a 90° RF pulse, the spin system is each time. Such a pulse sequence is called a satu-
saturated. ration recovery, and the resulting image is a
If repetition time (TR) is sufficiently long, all PD-weighted (PDW) image.
spins will relax to equilibrium. Such a pulse When shorter TRs are used, the amplitude of
sequence is called saturation recovery, but it is the second and subsequent FIDs is less than that
rarely used because it takes too long to make an of the first because enough time has not been
image. With short TR, the pulse sequence is allowed for full longitudinal relaxation. The
partial saturation. second and subsequent 90° RF pulses excite
The partial saturation pulse sequence is the already partially saturated proton spins, spins
simplest way to obtain weighted images. It con- that have not relaxed to equilibrium, such that
sists of a chain of 90° RF pulses (Figure 15-21). the FID will be of equal amplitude but less than
Before the first pulse, net magnetization (MZ) is that of the first FID. In such a partial saturation
at maximum value (M0), in equilibrium with the pulse sequence, the earliest FIDs are ignored.
external magnetic field (B0). A partial saturation pulse sequence is there-
The 90° RF pulse rotates the MZ onto the XY fore weighted by PD but also has a contribution
plane so that MZ is now zero and MXY is at a from T1, depending on the value of TR. Figure
maximum value determined by M0. Because MZ 15-22 shows the relaxation of MZ to equilibrium
equals zero, the proton spins are said to be satu- after a 90° RF pulse. Here, three tissues are con-
rated. They recover longitudinal magnetization sidered: cerebrospinal fluid (CSF), GM, and
according to the T1 relaxation time. Because M0 WM. CSF has the highest PD of the three; there-
is directly proportional to the PD, the initial fore its relative value of M0 is the highest. It
amplitude of the free induction decay (FID) should produce the brightest pixels.
depends on PD. CSF also has the longest T1 relaxation time,
For the proton spins to recover fully from approximately 1200 ms; it can be seen from the
saturation, a TR equal to at least 5 times the graph that CSF has the longest T1 because it rises
longest T1 is required. Clinically, this would from MZ equal zero with the lowest slope. There-
require a TR equal to many seconds, which fore CSF proton spins relax to equilibrium more
would result in unacceptably long imaging times. slowly than GM or WM proton spins. When
However, with a very long TR, the saturated equilibrium for all tissues is reached, however,
spins would fully recover, resulting in tissues M0 will be higher for CSF, resulting in brighter
with equal PDs having FIDs of equal amplitude pixels for CSF.
230 PART IV  Image Formation

M0 (CSF)

M0 (GM)

M0 (WM)

MZ
0 1000 2000 3000 4000 5000
TR (ms)
FIGURE 15-22  This shows relaxation of longitudinal magnetization (MZ ) during a partial saturation pulse sequence. A
more rapid relaxation to equilibrium (M0 ) occurs in tissues with short T1, such as white matter (WM). The amplitude of
M0 is a function of proton density, the gyromagnetic ratio, and B0.

GM has a lower M0 than CSF and a much At this 800 ms crossover, contrast reversal
shorter T1, approximately 1100 ms at 1.5 tesla. occurs. Below 600 ms, WM appears brighter
Therefore pixels representative of GM relax to than GM; above 800 ms, GM appears brighter
equilibrium more rapidly than CSF, but the equi- than WM. Tissue contrast is lost completely at
librium value is lower. the crossover. Furthermore, at a TR of less than
WM has the shortest T1 relaxation time of the approximately 2000 ms, CSF appears darker
three, approximately 650 ms at 1.5 T, and the than either GM or WM. This represents another
lowest PD. Therefore WM relaxes to equilibrium contrast reversal.
most quickly, but the equilibrium value is least Table 15-1 summarizes pixel appearance for
of the three. various ranges of TR. A partial saturation image
Figure 15-23 illustrates the necessity for select- made with a long TR is PDW; with a short TR,
ing the proper RF pulse sequence. If a long TR less than 500 ms, it is T1 weighted (T1W).
is selected, the image is PDW, and CSF is bright, Figure 15-23 shows a series of partial satura-
GM is gray, and WM is dark. tion images in which pixel character is a function
At a TR of approximately 3300 ms, CSF and of both PD and T1. There is little contrast
GM will appear equally bright—both brighter between the GM and WM in these images. These
than WM. If 90° RF pulses are applied at 3300 ms are not bad images in terms of signal-to-noise
intervals, the MZ of both CSF and GM will have ratio (SNR) and spatial resolution, but they are
recovered to the same value so that the represen- poor images in terms of contrast resolution and
tative FIDs have the same amplitude, and there visualization of brain anatomy. This lack of con-
is no difference in pixel intensity. Such a TR trast may seem peculiar because the pixel bright-
results in a crossover on the time scale of the T1 ness is a function of PD and T1, both of which
curves and an absence of contrast on the image. differ significantly between GM and WM.
Another crossover appears at approximately This is the problem with MRI pulse sequences;
2600 ms between CSF and WM and yet a third contrast information can be lost. Brain contrast
crossover at approximately 800 ms between GM due to PD can be viewed as making GM bright
and WM. The crossover at 800 ms is clinically and WM dark and the tissue contrast due to T1
significant because it is within the clinically as making GM dark and WM bright. The con-
acceptable TR of approximately 1000 ms, ensur- trast rendition is reversed between these two
ing reasonable imaging time. tissues.
 CHAPTER 15  Magnetic Resonance Images 231

FIGURE 15-23  This series of partial saturation images obtained at the indicated repetition times demonstrates contrast
reversal as a result of changes in proton density and T1 influence.

TA B L E 1 5 - 1 Pixel Appearance for Cerebrospinal Fluid, Gray Matter, and White Matter
as a Function of Repetition Time in a Partial Saturation or Inversion
Recovery Image

Tissue Repetition Time

Very Long Long Short
CSF Bright Dark Dark
GM Gray Bright Gray
WM Dark Gray Bright

Therefore when an image is made that puts partial saturation image obtained with an inter-
the two together, it is like doing a subtraction mediate TR.
film in angiography. Positive and negative The risk involved in doing routine MRI is that
images of the brain together result in a flat important underlying information may be lost.
image; contrast is lost. That is the case with the The MRI technologist must make sure that the
232 PART IV  Image Formation

TR
TI

RFt

RFs

M0

MZ

−M0 Null point

FIGURE 15-24  An inversion recovery radiofrequency (RF) pulse sequence consists of a 180° RF pulse followed by a 90°
RF pulse. The inversion time (TI) can be programmed for any time. Here it is shown occurring before the null point.

information lost is not important. This phenom- After shorter TIs, the FID amplitude is deter-
enon is not a function of imaging system design mined by the PD and T1 relaxation time because,
but a function of the MRI pulse sequence, which as in the partial saturation pulse sequence, the
is under the control of the MRI technologist. spins have not relaxed to equilibrium but, rather,
have remained partially saturated. Therefore
pixel brightness will be a function of PD and T1.
Inversion Recovery There is a value of TI where MZ equals zero.
A 180° RF pulse, rather than a 90° RF pulse, Application of a 90° RF pulse at this time results
is used to expand the range of longitudinal mag- in no signal because there is no Z magnetization
netization, MZ. The 180° RF pulse inverts the to rotate onto the XY plane. This value of TI is
MZ, which immediately begins to relax to equi- called a null point.
librium according to the T1 relaxation time
(Figure 15-24). At the null point, there is no transverse magne-
This relaxation of MZ is not detectable in the tization, hence no signal.
presence of the B0 field because it is too small.
Therefore an additional 90° RF pulse is needed The return to equilibrium magnetization after
to rotate the MZ onto the XY plane, where it will an inversion recovery pulse sequence follows a
be detectable. similar behavior as that from a partial saturation
This two-pulse sequence, a 180° RF pulse fol- pulse sequence (Figure 15-25). The principal dif-
lowed by a 90° RF pulse, is called an inversion ference is that the dynamic range is twice that for
recovery pulse sequence, and it is used to produce a partial saturation pulse sequence, resulting in
predominantly T1-weighted (T1W) images. The greater image contrast if the TI is properly
time between the initial 180° RF pulse and the chosen. At a TI of approximately 500 ms, the MZ
following 90° RF pulse is the inversion time (TI). of the three tissues is more widely separated than
If TI is very long, MZ will have relaxed to in a partial saturation pulse sequence.
equilibrium, and FID amplitude will be princi- There are some problems with the inversion
pally determined by PD. Such imaging times recovery pulse sequence. If the TI is chosen at a
would be even more objectionable than for the tissue null point, no signal is generated, and the
partial saturation sequence. overall SNR is poor. It is also important to
 CHAPTER 15  Magnetic Resonance Images 233

M0 (CSF)
M0 (GM)
M0 (WM)

MZ
2000 4000 6000 8000
TI (ms)

−M0 (WM)
−M0 (GM)
−M0 (CSF)
FIGURE 15-25  Relaxation of MZ during an inversion recovery pulse sequence covers twice the range of the partial satu-
ration pulse sequence. The result can be even greater image contrast.

M0 (CSF)
M0 (GM)
M0 (WM)

2000 4000 6000 8000

TI (ms)

−M0 (WM)
−M0 (GM)
−M0 (CSF)
FIGURE 15-26  If the magnetic resonance imaging system is incapable of displaying the negative range of MZ in the
inversion recovery, image information may be obscured by reduced dynamic range and confusing crossovers.

display negative values of MZ after the 180° RF the phase of the MR signal and the amplitude so
pulse in a manner that conveys unambiguous that contrast resolution following the scheme of
information. The most common approach is to Figure 15-26 is displayed.
display negative magnetization as black to gray, Inversion recovery images are usually high-
zero magnetization as pure gray, and positive contrast images that are T1 weighted. Repetition
magnetization values as gray to white. times in excess of approximately 3000 ms are
Rather than displaying the full dynamic range unacceptable because of low patient throughput.
of MR signals, some older imaging systems Inversion times of 300 to 600 ms are most often
display the negative magnetization as positive used. Such long TIs make multislice techniques
magnetization (Figure 15-26). This results in very time-consuming.
more crossovers for tissues and can confuse Several inversion recovery images are shown
image interpretation. To prevent this confusion, in Figure 15-27. These images show good con-
the imaging system must be capable of detecting trast between GM and WM. However, a closer
234 PART IV  Image Formation

A B

C D
FIGURE 15-27  This series of inversion recovery images shows the reversal of contrast and null regions as a function of
repetition time and inversion time.

look at the relative intensities of CSF, GM, and MR signal detected for imaging is usually a spin
WM as a function of TR and TI demonstrates echo, not an FID (see Chapter 14). Therefore a
contrast reversals and null regions. Careful pre- 180° RF pulse follows the 90° RF pulse in both
scription of the RF pulse sequence is required for sequences so that a spin echo is formed. This
inversion recovery images. does not significantly change the analysis of an
image if the time between the 90° RF pulse and
the spin echo is short.
Spin Echo The most widely used pulse sequence, the spin
The pulse sequences just described for partial echo pulse sequence, generates spin echoes at
saturation and inversion recovery imaging various times after the 90° RF pulse (Figure
sampled the primary MR signal, the FID. The 15-28). The 90° RF pulse rotates the MZ onto
 CHAPTER 15  Magnetic Resonance Images 235

TR

TE

RFt

RFs

M0

FIGURE 15-28  A spin echo radiofrequency (RF) pulse sequence showing the received magnetic resonance signals (RFs)
and the loss of transverse magnetization (MXY). The actual decrease in MXY is much faster than shown here because it
follows the T2* relaxation time (see Chapter 7).

M0 (CSF)

M0 (GM)

M0 (WM)

MXY
100 200 300 400 500 600 700 800 900 1000
TE (ms)
FIGURE 15-29  T2 relaxation curves for cerebrospinal fluid (CSF), gray matter (GM), and white matter (WM) are deter-
mined by the amplitude of MXY as a function of echo time (TE).

the XY plane just as in a partial saturation pulse overall SNR is reduced with later echoes. The
sequence. The FID that is formed is ignored, and relaxation curves at the bottom of Figure 15-29
sometime later, at one half the echo time (TE), a plot the loss of MXY, the transverse magnetiza-
180° RF pulse is applied. tion, and therefore, T2 relaxation.
This 180° RF pulse is called a refocusing pulse Partial saturation and inversion recovery
because it causes the proton spins to rephase (see images are principally PDW and T1W. Spin echo
Chapter 5). The rephased spins produce a spin images can be PDW, T1W, or T2W, depending
echo with amplitude less than that of the FID on the values of TR and TE. All images have a
because of spin-spin interactions and the result- contribution from PD.
ing loss of transverse magnetization due to the Figure 15-29 illustrates the T2 relaxation for
T2 relaxation in the tissue. the three tissues CSF, GM, and WM. As with the
Multiple spin echoes can be formed by mul- other pulse sequences, the initial signal intensity
tiple 180° RF pulses. Each spin echo is reduced from CSF, GM, and WM is determined by the
in amplitude by T2 relaxation; therefore the equilibrium magnetization, M0, for each. Loss of
236 PART IV  Image Formation

M0 (CSF)
M0 (GM)
M0 (WM)

MZ MXY
0 1000 2000 3000 4000 5000 10,000 500 1000
FIGURE 15-30  Relaxation of MZ to equilibrium and relaxation of MXY for an extremely long repetition time. If the image
is formed after a short echo time (TE), it will be weighted by proton density. Use of long TE results in T2-weighted images.

magnetization in the XY plane, MXY, and there- The initial signal amplitude is reduced for subse-
fore, MR signal intensity is determined by the T2 quent 180° refocusing pulses, and the relative
relaxation time of each tissue. amplitudes among the tissues also change.
Cerebrospinal fluid has a relatively long T2,
approximately 1500 ms. Gray matter and WM
have shorter T2 relaxation times, both approxi- If the TR is reduced to 1000 ms, GM exhibits
mately 100 ms; therefore the CSF signal remains a higher initial MR signal than CSF or WM
intense, resulting in a bright pixel for all TEs. (Figure 15-31). Therefore GM appears brightest
If TRs for a spin echo pulse sequence were on an early echo, and CSF appears brightest on
very long, the relationship between spin echoes a late echo. This is a contrast reversal resulting
for the three tissues would be like that shown in from a crossover.
Figure 15-30. CSF would always be brightest, If the TR is made very short, for example,
and WM would always be darkest because with 300 ms, there is little relaxation of MZ because
TR at least 5 times the longest Tl, all the proton of the T1 relaxation times being too long (Figure
spins have relaxed to equilibrium before each 15-32). WM now appears brightest on an early
new pulse sequence. echo. CSF is still brightest on a late echo. These
Figure 15-30 shows the relaxation of MZ back reversals of contrast result from crossovers of the
to equilibrium and the relaxation of MXY to zero T2 relaxation curves.
after a 90° RF pulse, both plotted on the same With a long TR and a long TE, the result
scale. It is clear that the relaxation of MXY is far is a T2W image. A long TR and a short TE
more rapid than the relaxation of MZ. result in a PDW image. Short TR and short
At a very long TR, for example, 5000 ms, MZ TE result in T1W images (Table 15-2). Short TE
has relaxed to equilibrium for nearly all tissues. produces early echoes, and long TE produces
Subsequent 90° RF pulses followed by 180° RF late echoes.
refocusing pulses result in images that are PDW All spin echo images have a PD contribution.
and T2W. At long TR, as the TE is reduced, the Depending on the T1 and T2 relaxation times of
image becomes more PDW. adjacent tissues, there will be many opportunities
If the TR is shortened, more Tl relaxation for obscuring contrast at crossovers.
influence is brought into the image because all A little thought shows that the amplitude of
tissues will not have recovered to M0. A review the spin echo from any tissue increases as the TR
of Figure 15-30 shows how magnetization among is lengthened, as TE is shortened, as T1 decreases,
these tissues changes with reduced TR. and as T2 increases. Stated differently, in general,
 CHAPTER 15  Magnetic Resonance Images 237

M0 (CSF)

M0 (GM)

M0 (WM)

MZ MXY
500 1000 500 1000
TR (ms) TE (ms)
FIGURE 15-31  Use of an intermediate repetition time (TR) does not allow full regrowth of MZ. Images formed with
short echo time (TE) are proton density (PD)–weighted and T1-weighted; images formed with long TE are PD weighted,
T1 weighted, and T2 weighted.

M0 (CSF)

M0 (GM)

M0 (WM)

MZ MXY
0

0
0
0

0
0

00
0
10
10

30

20

40
50
30

10

TR (ms) TE (ms)
FIGURE 15-32  With extremely short repetition time (TR), little MZ is reformed, resulting in more T1-weighted images
at all clinically used echo times (TEs).

TA B L E 1 5 - 2 Weighting of Spin Echo Images as a Function of Repetition Times

and Echo Times

TR
Short (<300 ms) Intermediate (300-1000 ms) Long (>2000 ms)
Short TE (<40 ms) T1 weighted PD and T1 weighted PD weighted
Long TE (>80 ms) T1 and T2 weighted PD, T1, and T2 weighted T2 weighted

a long T1 means less signal at a given TR, and a results in a T1W image, and long TE (late echoes)
long T2 means more signal at a given TE. results in a T2W image. Both images also have
The signal intensity of successive spin echoes some contribution from PD.
decreases regardless of TR; therefore, the SNR The spin echo images shown in Figure 15-33
also decreases. Long TE results in a dim, noisy represent variations of this most common imaging
image. As a general rule, short TE (early echoes) sequence. When considering spin echo imaging,
238 PART IV  Image Formation

A B

C D
FIGURE 15-33  Spin echo images produced at various repetition times with a long echo time result in T2-weighted
images. (Courtesy Orlando Diaz-Daza, Houston, TX.)

the radiologist must recognize whether the image contrast. On these late spin echo images, there is
is from an “early” echo or a “late” echo. These at least some differentiation between GM and
images were made with four different TR times, WM. Consequently, some other parameter must
ranging from 2000 ms to 4000 ms, at a TE of contribute to contrast. The additional parameter
130 ms; they are all late spin echo images. Late is PD, which always influences pixel intensity.
spin echoes are viewed primarily as providing a Longer TR images are brighter because of the
T2W image because the character of a pixel is relaxation of all tissues to their equilibrium
primarily related to the T2 relaxation of the magnetization.
tissue. These images are examples that can be used
However, these images are considerably to begin planning clinically useful spin echo pulse
different from the pure T2 image discussed sequences. To view detail on normal or abnormal
earlier. The pure T2 image is dull and has little anatomy, the MRI technologist would not want
 A B

C D
FIGURE 15-34  Spin echo images produced at various repetition times with short echo time result in varying proton
density–weighted and T1-weighted images. (Courtesy Orlando Diaz-Daza, Houston, TX.)

to do heavily T2W images. Rather, a PDW or TR, approximately 300 ms, the image is princi-
T1W image would be produced. pally T1W. If the TE is made very short (zero
Figure 15-34 shows spin echo images pro- time), a partial saturation pulse sequence with a
duced at various TRs varying from 350 ms to 270° RF pulse results.
4000 ms with a short TE of 35 ms. At long TR, A pathologic condition usually differs signifi-
in excess of 1000 ms, the image is PDW. At short cantly with T2. Figure 15-35, C, shows a late
240 PART IV  Image Formation

A B

C
FIGURE 15-35  The late spin echo image seen in C is most useful for identifying pathologic conditions, such as this
brain stem glioma. (Courtesy Orlando Diaz-Daza, Houston, TX.)

spin echo image that does not have much normal

CHALLENGE QUESTIONS
brain contrast, but the lesion is very distinct. The
contrast between normal tissues and pathologic 1. Define the term visual acuity. What is its
tissues of the brain is usually due to differences relationship to viewing an MR image?
in T2 relaxation. A heavily PDW or T1W image 2. When a spin echo image is made with a short
is nearly always needed to show correct anatomy, echo time and a long repetition time, what
and a heavily T2W image is used to show disease. weighting will the image exhibit?
 CHAPTER 15  Magnetic Resonance Images 241

3. MR images are said to be representational. 7. What is the principal advantage and disad-
What does this mean? vantage to inversion recovery imaging?
4. When you perform a double echo, spin echo 8. The spatial resolution of an MR image is
pulse sequence to produce two images, how determined by field of view and matrix size.
is each weighted? What determines spatial localization?
5. Which tissues have the longest relaxation 9. When the image of adjacent tissue loses con-
times? trast, how can the contrast resolution be
6. Medical images can be produced with trans- improved?
mitted, reflected, absorbed, or emitted elec- 10. What are the relative values for the relax-
tromagnetic radiation. Which of these terms ation times of soft tissues?
refers to MR images?
PA RT
V
Pulse Sequences
 CHAPTER

16 
Spin Echo Imaging

OBJECTIVES
At the completion of this chapter, the student • Identify the region of k-space most responsible
will be able to do the following: for contrast resolution and spatial resolution.
• Diagram a complete spin echo pulse sequence. • Diagram a fast spin echo pulse sequence.
• Diagram an inversion recovery pulse sequence. • Discuss the importance of echo train length.
• Define Hermitian symmetry.

OUTLINE
Spin Echo Pulse Sequence Contrast Enhancement with
Inversion Recovery Imaging Fast Spin Echo
Fast Spin Echo Imaging Stimulated Echoes
Partial Fourier Imaging J-Coupling
Rapid Acquired Relaxation Clinical Applications of
Enhanced Imaging Fast Spin Echo
Ordering of k-Space

KEY TERMS
Frequency encoding J-coupling Stimulated echoes
Half Fourier imaging Repetition time Temporal resolution
Hermitian symmetry Spoiler pulses

Since the introduction in 1980 of clinically resonance angiography [MRA]). Figure 16-1 is a
practical two-dimensional magnetic resonance timeline representing the increase in temporal
imaging (MRI) followed shortly by multi-echo, resolution of MRI since clinical introduction.
multislice imaging, magnetic resonance (MR) Various tissue motion and MRI techniques are
image quality has continued to improve. In addi- included on this timeline, along with the approx-
tion to better images, developments in MRI have imate dates of application of these techniques.
been driven by the goal of reduced image acquisi- MRI development was predicted to follow a
tion time. course similar to that of computed tomography
For instance, such reduced imaging time is (CT). An exponential introduction of new
required to image in the presence of normal imaging techniques and clinical capabilities were
tissue motion, including flowing blood (magnetic expected up to a saturation point where the

243
244 PART V  Pulse Sequences

0.001 2000

fMRI

Single 1996
EPI

Action
potential

0.01 GRASE 1992

MPRAGE

0.1 1990
RARE
Blink
1984 Snapshot
nds)

TURBO
eco

TIME

Cardiac Multi 1988

1.0 EPI
N (s

Cerebral
DAR
TIO

blood flow
LU

Swallowing
LE
SO

Respiration
CA

10 FSE 1986
RE

Peristalsis FLASH/GRASS
AL

Bolus passage
OR
MP

Breath-hold
100
TE

Multislice
Multiecho

1000 1982
CSE
Patient
10,000 tolerance Sensitive 1980
point
FIGURE 16-1  Timeline representing the continuing development of magnetic resonance imaging techniques and
required temporal resolution for imaging various physiologic motions.
 CHAPTER 16  Spin Echo Imaging 245

process would routinely remain clinical. Many

TABLE 16-1 Acronyms Used in Spin
predicted the death of CT because of the promise
Echo and Gradient
of MRI; however, MRI and CT are complemen- Echo Imaging
tary in many respects. Advances in CT such as
multislice spiral CT and computed tomography Acronym Explanation
angiography (CTA) have extended the capabili-
BASG Balanced SARGE
ties of CT. CE-FFE-T1 Contrast-enhanced T1-weighted FFE
MRI appears to be developing exponentially, CE-FFE-T2 Contrast-enhanced T2-weighted FFE
not toward saturation but rather ever increasing. CISS Constructive interference in the steady
This chapter and the following five chapters state
describe the enormous advances in MRI tempo- FAST Fourier acquired steady state
ral resolution. The term temporal resolution FE Field echo
refers to resolution in time in the same way that bFFE Balanced FFE
spatial resolution refers to resolution in space FFE Fast field echo
FGR Fast GRASS
and contrast resolution refers to resolution of
FIESTA Fast imaging employing steady state
one soft tissue from another. acquisition
FISP Fast imaging with steady state precision
Improved temporal resolution requires faster FLAIR Fluid attenuated inversion recovery
imaging. FLASH Fast low-angle shot
FSE Fast spin echo
Although many of these MRI advances have GFE Gradient field echo
resulted in faster imaging and increased patient GRASS Gradient recalled acquisition in the
throughput, they have also added new diagnostic steady state
capabilities. In addition to the imaging of cyclic HASTE Half Fourier acquired single shot turbo
spin echo
motion such as respiration and heartbeat, non-
HFI Half Fourier imaging
cyclic physiologic processes, such as peristalsis, IR Inversion recovery
saccadic eye motion, and joint movement, are MPGR Multiplanar GRASS
being imaged with MRI. Furthermore, MR fluo- MPRAGE Magnetization prepared rapid gradient
roscopy (MRF) is now possible. echo
Even the first paper on nuclear magnetic reso- PFI Partial flip angle
nance (NMR) by Felix Bloch in 1946 considered PSIF Reversed FISP
basic concepts leading to fast imaging techniques. RARE Rapid acquisition by repeated echo
The current development of fast and ultrafast RASE Rapid acquired spin echo
RF-FAST RF-spoiled FAST
imaging has been accompanied by an abundance
SARGE Spoiled steady state acquisition rewinded
of acronyms by vendors (Table 16-1). Each tech- gradient echo
nique is a development of spin echo (SE) or gra- SPGR Spoiled GRASS
dient echo (GRE) imaging. SSFP Steady state free precession
A rare, two-stalk century plant found in the STIR Short time inversion recovery
Big Bend region of west Texas is shown in Figure TFE Turbo field echo
16-2. The root spines represent MR image recon- True FISP Balanced SSFP
struction methods and the flowering pods repre- True SSFP Balanced SSFP
sent imaging techniques. TSE Turbo spin echo
TurboFLASH Version of FLASH with IR spin
preparation
SPIN ECHO PULSE SEQUENCE
The most commonly used MRI technique is
based on the SE pulse sequence that was
246 PART V  Pulse Sequences

FSE GRASE IFE

GRASE
TSE FGR
TURBO
FASE TGSE
RARE FLASH
FLARE
3D MP-RAGE FLASH

HFI QUEST SPGR

MPGR
BURST
DEFT SNAPSHOT FISP
GRASS FFE GRASS
RASE STEAM
FAST FEER
FATE PSIF

nt Echo
IRM IR CE-FAST

Gradie
ABSIR Spin Echo

k-space
MIP PC
T2 T2*
TOF 3DFT
2DFT

Image
Reconstruction

FIGURE 16-2  Magnetic resonance imaging techniques grouped according to the two principal schemes of image
acquisition.

introduced in 1950 by Hahn. This was first The terms radiofrequency (RF) pulse sequence
applied to imaging in 1980 and was quickly fol- and pulse sequence are often used interchange-
lowed by additional modifications in the pulse ably. However, RF pulse sequence correctly
sequence to allow multislice imaging. The Carr- applies only to the first line of the musical score.
Purcell-Meiboom-Gill (CPMG) pulse sequence is Therefore the term pulse sequence is used here
an improved SE pulse sequence for producing to represent not only the RF pulses but also the
images from multiple echoes. pulsed gradient magnetic fields and the received
 CHAPTER 16  Spin Echo Imaging 247

TR
TE
180°
90° 90°
RFt

GSS
1/2 TE

RFS

Gφ

GR
FIGURE 16-3  The basic spin echo pulse sequence showing the relative timing of each radiofrequency (RF) and gradient
pulse.

MR signals. The basic SE pulse sequence is RFt RFt

shown in Figure 16-3.
Often the lines of the musical score are identi- GZ BSS
fied as RFt, GZ, RFS, GY(ф), and GX(ω) (Figure
16-4, A). These symbols identify images that are RFS RFS
acquired in the transverse plane. Other orthogo-
nal and oblique planes are now frequently GY(φ) Bφ
obtained, so the symbols GZ, GY, and GX are less
meaningful. GX(ω) BR
In this and successive chapters, the following
A B
symbols are used: RFt, GSS, RFs, Bф, and GR
(Figure 16-4, B). In practice, any of the physical FIGURE 16-4  Musical score symbol (A) used in Chapters
1 through 17 to describe transverse plane images and 
(x, y, or z) gradient coils can perform the func- (B) generic form used in the rest of the text.
tions of slice selection, phase encoding, or
freqeuncy encoding. Thus, these symbols are
more appropriate for indicating the functional Sometimes GR is identified as the frequency-
utility of gradient magnetic fields. encoding gradient magnetic field.
The slice selection gradient is represented The GSS is energized to establish a difference
by GSS, which can define any of the three orthog- in resonance frequency along one axis. Usually,
onal planes. Oblique images are obtained when it is the long axis of the patient lying supine if
GSS represents any combination of the three transverse images are required. An RF pulse cov-
gradients energized simultaneously. The phase- ering a specified frequency range excites proton
encoding gradient magnetic field is Gф. The gra- spins in a slice of the patient.
dient magnetic field that is energized while a The transverse magnetization of the patient
signal is acquired is GR, the read gradient. (MXY) is then phase encoded, using Gф, in a
248 PART V  Pulse Sequences

Bφ

FIGURE 16-6  Phase-encoding with the Gф gradient mag-

netic field.
GR

FIGURE 16-5  Frequency encoding with the GR gradient solution, as discussed in Chapters 12 through 14,
magnetic field.
requires that this pulse sequence be repeated with
a number of different phase-encoding amplitudes
direction perpendicular to the long axis of the equal to the required matrix size.
patient, along either the anterior-posterior axis During the time between the 90° and 180° RF
(Y) or the lateral axis (X). This phase encoding pulses, the spin-spin or transverse relaxation
is accomplished by briefly energizing either the causes a dephasing of the transverse magnetiza-
X or Y gradient coils. When the phase-encoding tion. Other mechanisms, such as magnetic field
gradient magnetic field is off, all spins precess inhomogeneities, varying magnetic susceptibility
with the same frequency but now have different of different tissues, and chemical shift, also con-
phase shifts that depend on position along the tribute to the dephasing of MXY during this time.
phase-encoding gradient. The 180° RF refocusing pulse inverts the
phase of the spins by putting the faster compo-
Frequency encoding of the MRI signal helps nent of the transverse magnetization behind
locate that signal in the patient. the slower. This effect should persist over time,
except the dephasing caused by nonrandom pro-
Next, the remaining gradient magnetic field, cesses, such as static magnetic field, B0 inhomo-
GR, is switched on for a well-defined interval, geneity. Inhomogeneity is refocused after the
during which time the MR signal is received. 180° RF pulse to form the SE.
During this time interval, a difference in reso-
nance frequencies along the direction of fre- When refocusing is provided by a 180° RF pulse,
quency encoding occurs, causing the transverse the pulse sequence is called a spin echo pulse
magnetization to rotate with different frequency sequence.
depending on location along that axis (Figure
16-5). This is often termed frequency encoding. In SE imaging, the spatial frequency domain
The commonly used reconstruction algorithms (k-space) is filled line by line in sequential order
can distinguish between phase positions of oppo- from bottom to top (Figure 16-7). This process
site sign. The phase-encoding gradient has to be normally starts with large, negative phase-
of a defined amplitude and time duration to dis- encoding gradient amplitudes going through
tinguish between adjacent voxels (Figure 16-6). zero phase-encoding gradient amplitude and
Unfortunately, with just one phase gradient then to a large, positive phase-encoding gradient
amplitude, an ambiguous signal is received. The amplitude.
 CHAPTER 16  Spin Echo Imaging 249

128 KX (frequency)

256

Bφ KY (phase)
FIGURE 16-7  The ordering of k-space is sequential in spin echo.

Weak phase-encoding gradients provide infor- The measured lines of k-space associated with
mation about the coarse structure of the object strong phase-encoding gradient pulses contain
in the direction of phase-encoding contrast reso- the information about the smaller details (Figure
lution. Strong phase-encoding gradients provide 16-9). When the entire k-space is filled (Figure
information about smaller structures, spatial 16-10), a complete MR image can be recon-
resolution. structed with an inverse Fourier transform (FT−1).
The data are sorted into a raw data matrix. A significant source of artifacts in MR images
Each measured line of k-space consists of the is the variation in the subject from one phase-
analog SE sampled and converted to digital form encoded measurement to the next. The FT
during the read period. assumes that each line of k-space came from the
This read period is the time when the same slice of tissue in the subject. If the subject
frequency-encoding gradient is turned on and changes from one repetition time (TR) to the
data are acquired. The received signal contains next, this assumption is violated; therefore image
the phase information encoded before the read artifacts result.
period. Therefore the central zone of the k- If any object moves from one TR to the next
space contains the coarse structure of the object (blood flow, respiration, peristalsis, cardiac), an
(Figure 16-8). artifact may result. Phase alterations resulting
250 PART V  Pulse Sequences

126

127

128 KX (frequency)

129

130

Bφ KY (phase)
FIGURE 16-8  Low spatial frequencies acquired from the phase-encoding steps of weaker gradient pulses produce a
higher signal-to-noise ratio and higher contrast.

from motion while gradients are on produce magnetization of various tissues can be larger in
ghosting artifacts. IR imaging than in SE imaging (Figure 16-12).
The increased difference in longitudinal magne-
tization in the IR pulse sequence can therefore be
INVERSION RECOVERY IMAGING used to increase T1W and improve the contrast
Inversion recovery (IR) imaging was introduced between tissues with slightly different T1 values.
as an approach to enhance the T1-weighted
(T1W) contrast in SE imaging. For the con­ TI is the time between the 180° RF inversion
ventional IR technique, the longitudinal magne- pulse and the subsequent 90° RF pulse.
tization, MZ, is inverted before sampling a line
in k-space. The pulse sequence is shown in We note here that the use of the 180° RF
Figure 16-11. inversion pulse in the IR pulse sequence is just
After relaxation with the characteristic T1 one of the common spin preparation schemes
relaxation time for each tissue, the SE pulse that can be used to adjust image contrast by
sequence starts following a given inversion time manipulating the state of the magnetic nuclei
(TI). The differences among the longitudinal before the application of the excitation RF pulse.
 CHAPTER 16  Spin Echo Imaging 251

1
2
3

KX (frequency)

254
255
256

BF KY (phase)
FIGURE 16-9  High spatial frequencies acquired with strong gradient pulses produce noisier images with better spatial
resolution.

Other common methods used include spatial pre- Then, for a tissue with a short T1 relaxation
saturation, fat presaturation, and magnetization time, such as fat, an appropriately chosen short
transfer presaturation, which are described in TI allows elimination of the high signal from fat,
detail in the next chapter. thereby improving the visualization of true
A valuable application of IR for some clinical abnormalities in the regions of the image sur-
situations is the nulling of signal from tissues rounded by fat. Such pulse sequences are called
having a specific T1 relaxation time. If the SE short time inversion recovery (STIR) sequences.
part of the IR pulse sequence is started so that Figure 16-13 is a coronal T1W, fat-suppressed
the longitudinal magnetization of a relaxing image that shows right-sided optic neuritis.
tissue is crossing the zero line (MZ = 0) at the Fluids, such as cerebrospinal fluid, have a rela-
time when the 90° RF pulse is applied, there is tively long T1s, and TIs of 2 seconds or more are
no net MZ magnetization to be flipped onto the necessary to minimize the signal from fluid struc-
XY plane. Therefore no signal can be produced tures. Such a pulse sequence is called fluid attenu-
at that time for that tissue. If the T1 of the tissue ated inversion recovery (FLAIR). This allows a
is known, then the time TI(0) at which the MZ = better diagnosis of periventricular lesions in
0 can be determined to a good approximation which the bright pathologic areas are close to the
from the equation: TI = 1.44* T1. equally bright, fluid-filled spaces.
252 PART V  Pulse Sequences

1
2
3

128 KX (frequency)

254
255

256

KY (phase)
FIGURE 16-10  When k-space is filled, a magnetic resonance image with good contrast resolution and spatial resolution
can be reconstructed.

TR

TI TE
180° 180° 180°
90°
RFt

BSS

RFS

Bφ

BR
FIGURE 16-11  Inversion recovery spin echo pulse sequence.
 CHAPTER 16  Spin Echo Imaging 253

CSF M0

Brain M0

MZ

FIGURE 16-13  Fat suppression enhances the ability to

detect right-sided optic neuritis by reducing the high fat
signal from surrounding tissue. (Courtesy Robert Tien,
Singapore.)
SE

the Fourier transform produced using this method

presents the signal from the two tissues based
CSF M0
just on the magnitude of MXY, then there will be
no contrast between them.
Brain M0
A common technique used to resolve
this ambiguity considers the phase of the net
transverse magnetization, MXY, and is called
MZ
phase-sensitive reconstruction. Phase-sensitive IR
images present zero signal as an intermediate
gray in the image, whereas negative magnetiza-
tion is dark and positive values are bright.
Brain − M0
Phase-sensitive IR images also avoid the “zero
bounce-point artifact” that occurs when contrast
CSF − M0 is inverted between two tissues as a function
of TI.
IR
FIGURE 16-12  Longitudinal magnetization between FAST SPIN ECHO IMAGING
tissues is greater with inversion recovery (IR) than with spin
echo (SE). The time-consuming part of SE imaging is the
necessary repetition of GSS excitation and fre-
quency encoding GR with different Gф values to
STIR and FLAIR are pulse sequences designed to collect the spatial information of the slice. The
null the signal from fat and water, respectively. need to measure k-space line by line in SE imaging
is why as many as 256 to 512 repetitions are
IR techniques can lead to confusion if, for required.
instance, you want to distinguish between a long
T1 tissue that still has negative net magnetization
Partial Fourier Imaging
(−MZ) and a short T1 tissue that has a positive
net magnetization (MZ) at the TI time selected Two important statements about k-space and the
for a tissue with a medium T1 (Figure 16-14). If principle of phase encoding follow. First, when
254 PART V  Pulse Sequences

Short TI illustrated in Figure 16-15. Intuitively, it can be

seen that the MRI reconstruction schemes deal
Medium TI with redundant data. Theoretically, there is a
mathematically defined difference between a
negative phase-encoding gradient and a positive
phase-encoding gradient of the same amplitude.
Long TI
MZ
k-Space is symmetrical about both the X- (fre-
quency) and Y- (phase) axes.

The spatial frequency lines acquired on either

side of the zero amplitude phase-encoding gradi-
ent (Gф) are symmetrical. This mathematical rela-
IR
tionship is called Hermitian symmetry.
During imaging, this symmetry can be
destroyed in k-space by flow, RF phase shifts,
Short TI and T2 relaxation while the signal is sampled. In
the time domain, misadjusted quadrature signal
detection, shifts of the sampling grid, and gradi-
Medium TI
ent eddy currents that cause the echo to shift
during acquisition also cause the data to be
non-Hermitian.
Long TI The entire k-space is usually measured in SE
MZ
imaging to avoid artifacts within the image
caused by any violation of the assumption
of Hermitian symmetry. The redundant data
improves the signal-to-noise ratio (SNR) at the
cost of imaging time.
A number of approaches to fast imaging are
designed so that half the data are calculated
IRM
during image reconstruction by assuming Hermi-
FIGURE 16-14  Inversion recovery imaging has a contrast
range from −M0 to M0. Inversion recovery magnitude (IRM) tian symmetry, rather than being acquired during
considers only the absolute value of MZ and has a contrast signal detection. This technique is often termed
range from 0 to M0. half Fourier imaging (HFI). Using either half of
k-space (Figures 16-16 and 16-17) and calculat-
ing the other half are sufficient to produce an
switched on, the phase-encoding gradient mag- MR image.
netic field has to be sufficiently intense and long Such techniques have also been clinically
enough to establish a phase shift adequate to exploited, as in rapid acquired spin echo (RASE),
distinguish two adjacent voxels. Second, k-space but have become less important for the clinical
is filled by stepping from a maximum negative routine, even though they are slightly superior to
phase-encoding gradient amplitude through zero SE for certain applications.
to a maximum positive phase-encoding gradient Usually there are even better alternatives for
amplitude in succeeding TR intervals. those applications (see Chapters 17 and 18).
A 180° phase shift provided by a maxi- With half Fourier acquired single shot turbo spin
mum negative phase-encoding amplitude and a echo (HASTE), the HFI technique has another
maximum positive phase-encoding amplitude is useful clinical application. As for the SE method,
 CHAPTER 16  Spin Echo Imaging 255

KX (frequency)

Bφ KY (phase)
FIGURE 16-15  There is no difference in the spin echo obtained with gradients of equal amplitude but different
polarity.

only half of the data are sampled, and the other k-space. Hennig named his pulse sequence
half are then calculated. An additional eight lines rapid acquired relaxation enhanced (RARE)
are sampled for the phase correction to compen- technique.
sate for violations of Hermitian symmetry. The RARE technique at the time was geared
The lines of k-space to be measured are toward heavily T2W applications. The images
acquired from multiple SEs with a single exci­ were inferior in quality to SE images. Henning
tation. Figure 16-18 is a clinical example of produced his RARE examples on a scaled-down
a T2-weighted (T2W) study acquired within animal imaging system and was able to avoid
1 second with the HASTE approach. HASTE can eddy current effects that plagued fast MR imaging
be combined with inversion recovery magnitude of whole-body clinical scanners until the late
(HASTIRM) to null out fat (Figure 16-19). 1980s.
In 1989, Mulkern and Wong were working on
T2 measurements and needed a fast T2W local-
Rapid Acquired Relaxation izer. They adapted Hennig’s idea, recognizing the
Enhanced Imaging clinical potential of RARE. They reduced the
In 1986 Hennig suggested the use of multiple SEs heavy T2W and minimized the eddy current–
with phase encoding between each SE to measure related artifacts. Their technique is known as fast
256 PART V  Pulse Sequences

128 KX (frequency)

Bφ KY (phase)
FIGURE 16-16  Half Fourier imaging speeds image acquisition with little cost in signal-to-noise ratio.

spin echo (FSE), and a sample pulse sequence is FSE imaging uses multiple SEs within a TR, with
shown in Figure 16-20. This method has been each SE obtained by a different Gф.
adapted and refined by many vendors and mar-
keted with trade names, such as turbo spin echo
(TSE), dual echo fast acquisition interleaved spin Implementing FSE requires hardware changes
echo (DEFAISE), and contiguous slice fast spin for faster RF pulses and shielded gradients to
echo (CSFSE). allow acquisition of one SE every 10 to 15 ms
These FSE pulse sequences use multiple 180° with multiple RF excitations. However, the
refocusing RF pulses to produce a train of SEs required hardware changes are modest compared
within a single TR interval. With this increase with those required for echo planar imaging
in efficiency FSE has become the sequence of (EPI) in which SEs are obtained every millisecond
choice for T2-weighted spin echo imaging. A (see Chapter 20).
typical pulse sequence uses a TR of 4000 ms and Because multiple echoes are used to fill k-space,
effective TEs of 30 and 110 ms to produce a what echo time should be assigned to the image?
20-slice, double echo examination in less than Because of the importance of low spatial fre-
5 minutes. quencies and those phase-encoding steps with
 CHAPTER 16  Spin Echo Imaging 257

128 KX (frequency)

256

Bφ KY (phase)
FIGURE 16-17  Half Fourier imaging can be done equally well with data from either half of k-space.

FIGURE 16-18  A HASTE image, which is always T2 FIGURE 16-19  A HASTIRM (i.e., HASTE combined with
weighted, showing a hydrocephalus (one excitation, IRM) to null the fat signal (150 ms inversion time before
10.9 ms echo spacing, 87 ms effective echo time). (Cour- starting the HASTE sequence). (Courtesy William Faulkner,
tesy William Faulkner, Chattanooga, TN.) Chattanooga, TN.)
258 PART V  Pulse Sequences

TE1

TE2

TE3

TE4
90° 180° 180° 180° 180° 180° 90°
RFt

BSS

RFS

Bφ

BR
FIGURE 16-20  Signal relaxation during sampling of fast spin echo pulse sequence.

Effective echo time

90° 180° 180° 180° 180° 180° 180° 180° 90°
RFt

RFS

Bφ

FIGURE 16-21  Determination of effective echo time during fast spin echo imaging. The maximum signal occurs near
the center of k-space (maximum contrast resolution).

low gradient amplitudes, the echo time at which The effective echo time is that SE obtained
those Fourier lines are sampled is the important with the lowest amplitude Gф, usually the
one. The contrast at that time when the Gф = 0 middle SE.
will dominate the image contrast.
This time is called the effective echo time and
is used as the echo time assigned to the image After a first enthusiastic clinical evaluation of
(Figure 16-21). Spin echo trains as long as 64 are this technique, there came a period of caution
successfully used, but a 4- to 8-echo train is triggered by observation of differences in con-
typical. However, as the echo time is increased, trast between FSE and SE (e.g., bright fat)
the echoes become more heavily T2W, and arti- and by concerns of losing small objects in FSE
facts can be generated. imaging.
 CHAPTER 16  Spin Echo Imaging 259

The effect from the signal relaxation during

data sampling is similar to the effect observed
when a reduced matrix size is used. Depending
on the order of phase encoding, the spin echo
train length (ETL), and SE spacing, those spatial
frequencies acquired during very late echoes with
low signal contributions are underrepresented.
Acquiring high spatial frequencies with the
early SEs causes an overrepresentation, and edge-
enhancing artifacts may be observed. Acquiring
the high spatial frequencies with late SEs causes
underrepresentation, leading to blurring. Small
objects may be lost because the spatial resolution
is effectively decreased in the phase-encoding
direction. FIGURE 16-22  A periventricular lesion is seen on this
In FSE imaging, the signal for each subse- FSEIRM (fast spin echo inversion recovery magnitude)
quently measured line of k-space changes because image (T1, 2200 ms; repetition time, 9000 ms; effective
of T2. This is unlike SE imaging in which echo time, 150 ms). (Courtesy R. Nick Bryan, Philadelphia,
PA.)
the signal contribution is constant for each mea-
sured line. This condition is the cause of broad-
ening of the point spread function and is the
origin for the blurring sometimes observed in pulse sequences to study periventricular lesions
FSE imaging. (Figure 16-22).
The biggest concern in the clinical use of FSE The application of three-dimensional SE tech-
is the fear of not imaging small objects. The low niques has been developed. Using a higher spatial
spatial frequencies influence the contrast resolu- resolution than usually acquired with SE imaging
tion. The high spatial frequencies influence edge minimizes the concern of losing small objects.
definition and resolution of small objects. The argument is now quite the opposite; with
If high spatial frequencies are with the late FSE, better spatial resolution can be achieved in
SEs, they will be underrepresented because of less time than with SE. The result is SE applica-
the low signal intensity caused by the T2 relax- tions, previously thought to be prohibitively
ation. As with any other imaging technique, time-consuming, such as FLAIR-FSE and 3D-FSE,
pulse sequences and protocols have to be designed have become feasible.
to prevent the loss of small objects or small
lesions.
Ordering of k-Space
With FSE, superior spatial resolution with better The measured lines of greatest importance to
contrast resolution can be obtained in a reduced image contrast are those in which the low spatial
measurement time. frequencies are sampled. Low spatial frequencies
represent the coarse structure of the patient
Besides the potential reduction in measure- and determine the contrast of the image. Those
ment time, FSE has expanded clinical capabilities lines of k-space are sampled with low amplitude
that were otherwise too time-consuming, such phase-encoding gradient magnetic fields.
as the selection of a TR beyond 3 seconds If multiple lines are acquired with more than
and the acquisition of 512-pixel or even 1024- one SE, the SE following the weakest phase-
pixel matrices. TIs of several seconds have been encoding gradient pulse is called the zero order
applied to null out the signal from fluid in FLAIR spin echo. The sampled line for this zero order
260 PART V  Pulse Sequences

of pure water can be long: up to several seconds.

However, those water molecules that become
attached to a macromolecule through hydrogen
bonds can pass their energy on to the protons
within the macromolecule because of their rela-
tively fixed geometric relationships. This is the
dipole-dipole interaction, and it has a range of
several hundreds of micrometers.
Now, if the bound water protons come into
the vicinity of the free water protons, the free
water protons have a greater likelihood of giving
their energy up to the bound water protons. This
is due to these protons being attracted to the
slowly tumbling macromolecule, inducing them
FIGURE 16-23  Left side, fast spin echo. Right side, con- to stay in the neighborhood a bit longer than
ventional spin echo. Notice the hyperintense subcutane- other free water protons.
ous fat in the lower left of the image. (Courtesy R. Nick The bound water protons now act as a conduit
Bryan, Philadelphia, PA.) for the energy to be transferred from the free
water protons through the bound water protons
to the protons in the macromolecules. The bound
SE fills the zero order in k-space and defines the water acts like an energy pipe, and the macro-
effective echo time. Without changing the basic molecules act as an energy sink. More important,
pulse sequence, the weighting of the achieved the net result of this improved energy transfer
image is changed according to the effective echo mechanism is the decrease of T1 from what
time by just reordering the measurement of it would be if there were no macromolecules
k-space. present.
When an off-resonance MTS RF pulse is
applied to the bound water, the bound water
CONTRAST ENHANCEMENT protons receive all the energy they can handle.
WITH FAST SPIN ECHO
These protons are now saturated and will not be
The differences between the contrast achieved able to pass off any of the energy from the free
with FSE and the contrast observed in SE imaging water protons until they get rid of the energy
are most noticeable in fat (Figure 16-23), in from the MTS RF pulse. This blocking of the
delineation between gray and white matter, and energy transfer pathway effectively increases T1
in a reduced sensitivity to magnetic susceptibility for tissues that have a large concentration of
gradients with FSE. The better delineation of macromolecules and makes them appear darker
gray and white matter can be explained as a on the MR image.
magnetization transfer saturation (MTS) phe-
nomenon (see Chapter 17).
When blood is moving, it is not affected by MTS,
First, let us consider the major relaxation
and therefore this process improves contrast on
mechanism of protons in tissue. This is dipolar
MR angiograms.
coupling. However, dipolar coupling between
protons in water molecules does not happen
readily because these molecules are tumbling In FSE, multiple 180° RF pulses are used for
rapidly about and rarely interact. refocusing the spins into an SE. The RF pulse
Also, the free water protons have no other for one slice is off-resonance for the remaining
spins to pass their excess energy to. Thus the T1 slices, causing a saturation of the short T2
 CHAPTER 16  Spin Echo Imaging 261

components. By means of magnetization transfer, before they become a problem for clinical FSE
this leads to increased saturation of the free images.
water.
This saturation causes a further improvement
in contrast and is, to a certain degree, responsible
J-Coupling
for the better contrast between gray and white The term J-coupling helps explain the relatively
matter in FSE imaging. The resulting FSE con- bright fat observed in T2W FSE imaging (see
trast is true T2 contrast, rather than the T2* Chapter 9). For complicated molecular struc-
contrast associated with GRE imaging (see tures like fat, there is a coupling mechanism
Chapter 18). between spins that forces the MXY magnetiza-
tion to more rapidly dephase despite the 180°
refocusing RF pulse. This dephasing superim-
Stimulated Echoes posed on the T2 relaxation results in the addi-
Another significant difference between FSE and tional signal loss in fatty tissues seen in SE
SE is the generation of stimulated echoes. The imaging.
180° refocusing RF pulses in the FSE sequence For a rapid sequence of 180° refocusing RF
are generated much more closely together in pulses, it can be shown that this J-coupling
time than those in single-echo or dual-echo SE pattern is broken. The missing J-coupling dephas-
sequences. Because of imperfections in the 180° ing means there is relatively more signal from
RF pulse some of the Mz magnetization is not lipids, causing the bright fat on FSE imaging.
refocused, but remains aligned along the negative
Z-axis. This component of net magnetization is
Clinical Applications of Fast
affected by a subsequent 180° refocusing RF
Spin Echo
pulse, producing an additional echo called a
stimulated echo. The imperfection of the 180° Fast spin echo has replaced SE as the T2W pulse
refocusing RF pulse has its origin within the side sequence of choice for most clinical applications.
lobes of the slice profile where the RF pulse is For most examinations, the time saved with the
less than 180°. use of multiple echoes per TR is less often used
Stimulated echoes can be produced any time to reduce measurement time than to increase
three or more RF pulses are applied in rapid suc- spatial resolution. Increasing matrix size can be
cession; that is if the time between the pulses is used to improve the spatial resolution.
less than T2*. In GRE imaging (see Chapter 18), With the use of a rectangular field of view in
these so-called stimulated echoes are eliminated conjunction with FSE techniques, imaging of the
with gradient spoiler pulses or RF spoiling by spine is done faster with a significantly increased
phase cycling. In FSE, these stimulated echoes are spatial resolution compared with SE imaging
managed by using crusher gradients, applied on techniques. Figure 16-24 shows a T1W cervical
either side of the 180° refocusing RF pulses. If spine study acquired in 3 minutes, 25 seconds,
the stimulated echoes are not eliminated with the with a 512 matrix.
crusher gradient pulses then ghosting artifacts For the lumbar spine study shown in Figure
will appear, which resemble those caused by 16-25, the time savings of the FSE technique is
patient motion or system instabilities. Stimulated even more obvious. A T2W, 512 matrix image is
echoes produce aritfacts that have subtle differ- obtained in 1 minute, 56 seconds.
ences in their phase characteristics, compared to For abdominal applications, T2W measure-
motion and other sources of MR image ghosting. ments can now be acquired in a breathhold, as
Fortunately, these ghosts can be identified and shown in the 512-matrix image in Figure 16-26.
diagnosed in phantom images by a medical phys- FSE techniques are successfully applied to pelvic
icist and corrected by the MRI service engineer and orthopedic studies as well.
262 PART V  Pulse Sequences

FIGURE 16-24  T2W FSE study of the cervical spine (TR, FIGURE 16-26  T2-weighted fast spin echo of the liver
3250 ms; effective echo time, 112 ms). (Courtesy Pedro acquired in only 18 seconds (repetition time, 1240 ms;
Diaz-Marchan, Houston, TX.) effective echo time, 120 ms). (Courtesy Hani Haykal,
Houston, TX.)

CHALLENGE QUESTIONS
1. For two-dimensional Fourier transform
(2DFT) MRI, which gradient coil is energized
to produce a given slice?
2. What are some advantages to FSE imaging
over SE imaging?
3. What is the purpose of the 180° RF refocusing
pulse in SE imaging?
4. Describe the difference in the pulse sequence
between SE and FSE.
5. What does the second FT refer to in 2DFT
MRI?
6. What do STIR and FLAIR relate to?
7. What is a stimulated echo?
8. What is meant by Hermitian symmetry?
9. What is effective echo time?
FIGURE 16-25  T2-weighted fast spin echo study of the
lumbar spine acquired in 2 minutes (repetition time,
4000 ms; effective echo time, 120 ms). (Courtesy Pedro
Diaz-Marchan, Houston, TX.)
 CHAPTER

17 
Chemical Shift and
Magnetization Transfer

OBJECTIVES
At the completion of this chapter, the student • Calculate pixel shift for any B0 field and any
should be able to do the following: receiver that samples bandwidth.
• Differentiate among free water protons, fat • Define magnetization transfer (MT).
protons, macromolecular protons, and bound • Discuss the proton spins involved in MT.
water protons. • Draw an MT, fat saturation, and inversion
• Calculate the water/fat chemical shift for any recovery pulse sequence.
B0 field.

OUTLINE
Chemical Shift Magnetization Transfer

KEY TERMS
Bound water protons Parts per million
Free protons Sampling bandwidth

The gyromagnetic ratio for hydrogen is situation applies only to proton spins in mobile
42 MHz/T. Therefore, in a 1 T B0 field, hydrogen water molecules, so-called free water.
nuclei precess with a resonant frequency of Although water makes up approximately
42 MHz. Table 17-1 summarizes hydrogen 80% of all tissue molecules, hydrogen is only
proton resonant frequency at several other B0 approximately 60% of all tissue atoms. Never-
values. theless, the hydrogen in water predominates and
For a magnetic resonance imaging (MRI) is the principal source of the magnetic resonance
signal to be obtained from tissue, the transmitted (MR) signal.
radiofrequency (RFt) must be at the prescribed
Sixty percent of the atoms of human tissue are
resonant frequency, say 42 MHz for a 1 T MRI
hydrogen atoms.
system, to excite the proton spins. However, this

263
264 PART V  Pulse Sequences

TA B L E 1 7 - 1 Resonant Frequency for Hydrogen Protons as a Function of Magnetic Field

Strength B0

B0 0.2 T 0.35 T 0.5 T 1.0 T 1.5 T 3.0 T 4.0 T

Resonant frequency 8 MHz 15 MHz 21 MHz 42 MHz 63 MHz 126 MHz 168 MHz

water spins protein spins an arbitrary value of 0 on the ppm scale. Water
protons are shifted downfield 4.7 ppm and those
of fat, about 1.2 ppm. This is shown in the spec-
trum of Figure 17-2.
The reason that the chemical shift is expressed
in ppm is that the shift will increase as B0
increases. Thus, in a 1-T B0 field, water protons
will resonate at 300 Hz lower than TMS and fat
protons have a resonant frequency about 51 Hz
below that of TMS. However in a 3 T B0 field,
fat spins water protons will resonate at 900 Hz lower
than TMS and fat protons will resonate about
153 Hz below TMS. Using the ppm scale allows
the chemical shift to be more easily related
between MRI systems with different B0 fields.
All hydrogen in water is bound to oxygen in
the same manner. Therefore the water peak is
narrow and intense. Hydrogen in fat is bound in
many different ways, and this causes the fat peak
FIGURE 17-1  The three-compartment model of proton to be broader and less intense. It is actually the
spins in the patient. result of many individual spectral lines, each rep-
resenting a different type of chemical bond.
However, proton spins in other tissue atoms The resonant frequency for fat differs from
also contribute to the MR signal if they are that for water because of very small differences
excited at the proper resonant frequency. Many in the local magnetic field. Both molecules may
schemes have been proposed for compartmental- reside in an equal B0 field, but the local magnetic
izing body tissue for the purpose of MRI. The field is slightly different and determined by the
three-compartment model (Figure 17-1) is configuration of electrons in the bound molecule.
the simplest and all that is necessary for the This phenomonon is known as “electron shield-
discussion of magnetization transfer (MT)— ing.” Hydrogen spins and other nuclear spins
water protons, fat protons, and macromolecular in the same molecule can also cause smaller
protons. Before proceeding to MT, consider disturbances of the local magnetic fields, which
again chemical shift because they are related. is manifested in the J-coupling mentioned in
Chapter 16.
CHEMICAL SHIFT Free water protons and fat protons have a reso-
A revisit to the discussion of the parts per million nance frequency separation of about 3.5 ppm.
(ppm) scale in Chapter 7 will show that ppm is
a relative value compared with the standard mol- Because fat and water proton spins are at
ecule tetramethylsilane (TMS). TMS is assigned 4.7 and 1.2 ppm, respectively, their resonant
 CHAPTER 17  Chemical Shift and Magnetization Transfer 265

3.5

noise

8 7 6 5 4 3 2 1 0
ppm
FIGURE 17-2  The proton magnetic resonance spectrum, showing the separation of 3.5 ppm between fat and water.

frequency separation is (4.7 − 1.2 = 3.5) 3.5 ppm. If the receiver sampling bandwidth is, for
At 0.5 T, the frequency difference is small. instance, 8 kHz, the signal will be sampled
However, at 1.5 T and higher, the frequency dif- as shown in Figure 17-4, and at 1.5 T the pixel
ference can result in a significant artifact. shift is approximately 7. Increasing the sampling
bandwidth to 32 kHz increases the number of
Fat/Water Chemical Shift samples and reduces the chemical shift to less
than 2 pixels.
at 0.5 T: 3.5 ppm ¥ 0.5 T ¥ 42 MHz/T
= 74.5 Hz
at 1.5 T: 3.5 ppm ¥ 1.5 T ¥ 42 MHz/T
= 223.5 Hz
Sampling Bandwidth/Pixel Shift
at 3.0 T: 3.5 ppm ¥ 3 T ¥ 42 MHz/T
= 447 Hz 8 kHz
= 31 Hz/pixel
56 Pixels
Consider the water-filled cyst in the fat glob 73.5 
of the patient in Figure 17-3. In conventional = 2 pixel @ 0.5 T, 
 31 
SE and GE imaging the direction of the chemi-
220 
cal shift is always along the direction of the = 7 pixel @ 1.5 T, 
frequency-encoded read gradient magnetic field,  31 
GR. The amplitude of the shift is proportional to 16 kHz
= 63 Hz/pixel
the external magnetic field, B0. The appearance 256 pixels
of the chemical shift as an artifact is determined 73.5 
by the receiver sampling bandwidth. = 1 pixel @ 0.5 T, 
 63 
For a 256 × 256 image matrix, there will be
220 
256 MR signals acquired with 256 different = 3 pixel @ 1.5 T, 
phase-encoding gradient magnetic fields, Gф.  63 
Each signal will be sampled and encoded to 256 32 kHz
= 125 Hz/pixel
pixels along the frequency axis. In k-space, the 256 pixels
frequency axis is KY. 73.5 
= 0.5 pixel @ 0.5 T, 
 125 
Receiver bandwidth is the frequency at which an = 2 pixel @ 1.5 T, (220/125)
MR signal is sampled.
266 PART V  Pulse Sequences

Fat glob

Fluid−filled cyst

GR

Fat Fluid

Enhanced signal

Fluid GR

Signal void
FIGURE 17-3  The chemical shift artifact appears
in the direction of the frequency-encoded read
gradient magnetic field, GR. Fat

MAGNETIZATION TRANSFER
Two methods are used to enhance image contrast
The proton MR spectrum of Figure 17-2 is not by accentuating the signal from water protons.
exactly complete. In addition to the resonance The use of inversion recovery to invert fat
peaks of water protons and fat protons, there is protons and selectively suppress their signal
a broad background of spins from protons in based on T1 was discussed in the previous
macromolecules and water bound to macromol- chapter. A 180° RF inversion pulse is followed
ecules (Figure 17-5). by a 90° pulse applied at an inversion time (TI),
Fat and water protons exhibit narrow reso- which just matches the time when fat longitudi-
nant peaks and relatively long T2—on the order nal magnetization is relaxing through zero (where
of tens of milliseconds. Macromolecular protons TI = 1.44 × T1) from −MZ to +MZ (Figure 17-6).
and their associated bound water protons exhibit The fat inversion pulse sequence followed by
a wide spectrum of resonant frequencies and a the appropriate TI interval before spin excitation
very short T2, less than one millisecond. removes those bright fatty tissues from the image
These characteristics of resonance and T2 and therefore improves the contrast of remaining
result in a transfer of magnetization from free tissue. However, more can be done by transfer-
water through bound water to macromolecular ring the magnetization from macromolecular
protons, resulting in reduction in image contrast. protons and associated bound water protons.
 CHAPTER 17  Chemical Shift and Magnetization Transfer 267

7 pixels @ 1.5 T

8 kHz

3 pixels @ 1.5 T

Sampling
frequency
16 kHz

2 pixels @ 1.5 T

32 kHz

FIGURE 17-4  Signal sampling bandwidth controls chemical shift artifact.

3.5 ppm

Water
Fat

Macromolecules

Noise

FIGURE 17-5  The broad spectrum of magnetic resonant frequencies of protons bound to macromolecular reduces
contrast by transfer of magnetization to water protons.
268 PART V  Pulse Sequences

Fluid

Fat

Muscle

TI

Mz
100 300 2000 3000
TR (ms)
Null

FIGURE 17-6  The signal from fat is nulled when the inversion time of an inversion recovery pulse sequence occurs at
the time fat MZ = 0.

The signal from water protons is relatively Figure 17-8 is an example of MT gradient echo
long, relatively intense, and of narrow frequency imaging. The details of the mechanism of MT
range. The signal from macromolecular protons are too complicated to go into here, but under-
and bound water protons is very short and of standing that MT involves the saturation and
low intensity and wide frequency range. However, therefore suppression of signal from macromo-
it is a signal that can be used to enhance the lecular and then bound water will give you a
contrast of free water protons in different tissues. basic picture of how it works. Incomplete fat
The signal from free protons can be increased saturation is demonstrated in Figure 17-9.
by saturating the bound spins with an RF pulse MT techniques are particularly helpful in
offset from the water resonance by a few kilo- magnetic resonance angiography (MRA) and
hertz (Figure 17-7). This RF pulse effectively contrast enhanced imaging. During MRA, MT
saturates the magnetization of bound water improves vessel contrast by suppressing sur-
protons so that they no longer act as a conduit rounding tissue. Gadolinium contrast imaging is
to transfer energy from free water protons to used similarly to enhance lesions that take up the
protons of molecules that are in a bound state. gadolinium while surrounding normal tissues,
The result is better contrast for fat/water protons. such as fibroglandular tissues in the breasts,
The MT saturation pulse is followed by any can be darkened at the same time using MT
number of routine imaging pulse sequences. contrast.
 CHAPTER 17  Chemical Shift and Magnetization Transfer 269

Free water

Bound water
Fat

Macromolecules

Noise

Off resonance
MT pulse

Free water

Fat

Macromolecules

Noise

FIGURE 17-7  With the application of a broadband RF pulse centered off the water/proton peak, the macromolecular
protons become saturated and magnetization is transferred to water.

TR

90°

RF

Off res

BSS

Bφ

BR

GRE
RFS

TE
FIGURE 17-8  The magnetization transfer pulse is an off-resonance pulse that precedes a normal imaging pulse sequence,
such as the gradient echo shown here.
270 PART V  Pulse Sequences

FIGURE 17-9  An example of incomplete fat satu-

ration due to magnetic field inhomogeneity. (Cour-
tesy Wlad Sobol, Birmingham, AL.)

CHALLENGE QUESTIONS 7. What is the general relationship between

receiver bandwidth and the chemical shift
1. What happens if the RF pulse programmer artifact?
is tuned to 70 MHz in a 1.5 T MRI system? 8. What is the chemical shift between fat and
2. New imaging systems are being installed water for a 512-image matrix sampled with
with a B0 = 3 T. What is the operating fre- a 16 kHz receiver?
quency for such an MRI system? 9. Along which gradient magnetic field—GSS,
3. Why is hydrogen used for MRI? Gф, or GR—would a chemical shift artifact
4. What is magnetization transfer (MT)? appear?
5. Which has lower resonant frequency: hydro- 10. What is meant by fat saturation?
gen in fat or hydrogen in water?
6. Relative to mobile water protons, at what
frequency do macromolecular protons and
bound water protons resonate?
 CHAPTER 17  Chemical Shift and Magnetization Transfer 271
 CHAPTER

18 
Steady State Gradient
Echo Imaging

OBJECTIVES
At the completion of this chapter, the student • Identify a steady state pulse sequence.
should be able to do the following: • Explain the meaning of a stimulated echo.
• Distinguish between a spin echo (SE) and a • Define Ernst angle.
gradient echo (GRE). • Discuss the contrast weighting of spoiled GRE
• Explain how a gradient echo is formed using a and refocused GRE.
bipolar read gradient.

OUTLINE
The Gradient Echo Spoiled GRE Versus Refocused
T2 Versus T2* GRE
Steady State Contrast-Enhanced Techniques
Low Flip Angle
Ernst Angle

KEY TERMS
Balanced steady state free Ernst angle Steady state
precession Low flip angle
Dephase Spoiled

In 1984, the desire for faster imaging led to magnetization (MZ), leading to a lower signal-to-
the introduction of a group of pulse sequences. noise ratio (SNR) and less contrast than if TR
These pulse sequences lack the 180° radiofre- were shortened while the flip angle remained
quency (RF) refocusing pulse used in spin echo at 90°.
(SE) imaging. The elimination of the 180° RF The solution to this problem was extracted
pulse allows the use of a shorter repetition time from nuclear magnetic resonance (NMR) spec-
(TR) than is possible with SE imaging. This troscopy in which the flip angle is not fixed at
results in a decrease in the available longitudinal 90°. Rather, it is optimized for the TR of the

272
 CHAPTER 18  Steady State Gradient Echo Imaging 273

α
RFt

BSS

RFs

Bφ

BR

FIGURE 18-1  The basic pulse sequence for gradient echo

imaging.

FIGURE 18-2  Following an α pulse all spins precess at

measurement and the spin-lattice relaxation time the same frequency with the same phase and dephase
(T1) of the sample. The use of excitation flip with the same T2*.
angles less than 90°, i.e., the use of an α pulse,
and the omission of the 180° RF refocusing pulse
have allowed adequate SNRs within shorter sam-
pling times.

In gradient echoes (GREs), spins are refocused

with a gradient magnetic field, rather than a
180° RF pulse.

THE GRADIENT ECHO

Bφ
Figure 18-1 shows the basic pulse sequence for
GRE imaging. Immediately after a low flip angle
excitation, α pulse, in the presence of a slice- FIGURE 18-3  When a pulsed phase-encoding gradient
select gradient magnetic field (GSS), a free induc- magnetic field is applied for an instant to the spins of
tion decay (FID) is generated. The negative lobe Figure 18-2, a phase change occurs.
of BR is a rephasing lobe. Spins are in phase and
precessing at the same frequency (Figure 18-2).
As in SE imaging, the phase-encoding gradient or after the phase-encoding gradient magnetic
magnetic field (Gф) is energized momentarily to field Gф, the spins dephase with T2* relaxation
produce a difference in phase along the Y-axis. time (Figure 18-4).
The spins in any column along the Y-axis precess The read gradient magnetic field, GR, causes
at the same frequency, but now a difference in a range of resonance frequencies during data
phase occurs (Figure 18-3). Still, each spin relaxes acquisition. This range in frequencies also
with equal T2* relaxation time. causes dephasing, which needs to be considered
If a frequency-encoding read gradient mag- and compensated. Providing GR with opposite
netic field (GR) is energized simultaneously with polarity and half the duration of that same
274 PART V  Pulse Sequences

gradient during the data acquisition causes an

accelerated dephasing; this is followed by rephas-
ing and the generation of a gradient-refocused
echo, which is known as a gradient echo or GRE
(Figure 18-5).
Another view of this method of echo forma-
tion is shown in Figure 18-6. The GRE is shown
here as forming within the FID. This is the prin-
cipal reason GRE can be so fast.

T2 VERSUS T2*
The GRE signal is generated from an FID when
the dephasing gradient is applied early in the FID
and immediately reversed. The GRE forms within
BR the time of the FID, and the amplitude depends
on T2*.
When an SE is used to generate a GRE, the
FIGURE 18-4  When a frequency-encoding gradient refocusing GR can be switched during the first
magnetic field is applied to the spins of Figure 18-3, T2*
relaxation is accelerated.
2 4 6 8 10

1 3 5 7 9

Spins

GR

1 2 3 4 5 6 7 8 9 10

Out of phase

In phase

Out of phase

RFs

FIGURE 18-5  A bipolar read gradient magnetic field, BR, instead of a 180° radiofrequency (RF) pulse, produces the
gradient echo.
 CHAPTER 18  Steady State Gradient Echo Imaging 275

α α

RFt RFt

FID GRE

RFs RFs

BR

B
FIGURE 18-6  A, When BR is energized during a free induction decay, spins dephase more rapidly. B, Reversing the
polarity of BR causes the spins to rephase, forming a gradient echo (GRE).

half of the SE (Figure 18-7, A) or after the center 18-9, A, the application of GR causes the spins
of the SE (Figure 18-7, B). In either case, the that were dephasing with T2* relaxation time to
amplitude of the GRE depends on T2*. Radiolo- dephase even more rapidly. If the polarity of GR
gists rarely use this technique because it requires is reversed, as in Figure 18-9, B, the accelerated
more time. dephasing is reversed and the spins rephase to
In addition to the gradient magnetic fields that form a GRE. The amplitude of this GRE is that
are energized in conjunction with a selective exci- of the FID at that time of relaxation.
tation α pulse for the purpose of spatial encod- Because they usually remain at the same loca-
ing, patient- or system-related factors cause spin tion and are constant over time, these dephasing
dephasing, which makes T2* still shorter. Inho- mechanisms are refocused with the 180° RF
mogeneity of the primary static magnetic field pulse used in SE imaging. The amplitude of the
(B0) and the magnetic susceptibility differences signal intensity in SE imaging depends on T2, not
among tissues are typical sources for such spin T2* (Figure 18-10, A).
dephasing. With GRE imaging, the additional dephasing
The influence of B0 inhomogeneity is further mechanisms contribute to the image contrast
described in Figure 18-8. The voxel in Figure and the apparent spin-spin relaxation time is
18-8, A, contains five spins aligned and in phase dominated by T2*, rather than T2 (Figure 18-10,
to produce transverse magnetization, MXY, pre- B). Figure 18-11 shows the dephasing effects of
cessing with frequency ω. These spins dephase magnetic susceptibility gradients at the base of
with time because of T2 relaxation. the skull.
In Figure 18-8, B, the B0 does not have uniform
intensity (magnetic field inhomogeneity) as indi-
cated by the small arrow. The result is acceler-
Steady State
ated relaxation (dephasing), which is T2* One of the first groups to develop GRE imaging
relaxation. called its technique fast low angle shot (FLASH).
Even faster relaxation will result if a gradient Other acronyms for this type of GRE pulse
magnetic field is applied to the voxel. In Figure sequences, such as SPGR, CE-FFE-T1, and
276 PART V  Pulse Sequences

90° 180°
RFt

SE

RFs

RFs

BR

RFs

BR

B
FIGURE 18-7  A gradient echo can be formed either early or late within a spin echo (SE).
 CHAPTER 18  Steady State Gradient Echo Imaging 277

t=0 t=1 t=2 t=3 t=4

ω ω ω ω ω
MXY

t=0 t=1 t=2 t=3 t=4

ω ω ω ω ω
MXY

B
FIGURE 18-8  A, Spins randomly dephase because of T2 relaxation. B, Dephasing is more rapid in the presence of a
gradient magnetic field as a result of reversible magnetic field inhomogeneities.

GFE, have been used by various MRI system amount of MZ reduced because of the low flip
manufacturers. angle excitation.
FLASH is the prototype of a class of GRE
pulse sequences, called poiled GRE, which uses Steady state is the condition of constant longitu-
a steady state for the longitudinal component of dinal magnetization after repeated α pulses.
the macroscopic magnetization, MZ, to produce
a T1W signal that is controlled by the TR and In other words, if the decrease in MZ because
the α excitation RF pulse (Figure 18-12). Three of the excitation is larger than the amount recov-
repetitions from a long train of low flip angle α ered during the TR, the MZ becomes smaller. A
pulses are shown in Figure 18-13. Each α pulse point is reached at which the reduction of mag-
generates an FID, as well as SEs. The SEs occur netization caused by the excitation equals the
because of the rephasing of spins from the previ- same amount as the recovery during TR. That
ous α pulses. The total signal, RFs, is the sum of condition is steady state (Figure 18-14).
the FID and the SE. If the longitudinal magnetization is far from
The further away from the fully relaxed its fully relaxed state of equilibrium, the absolute
state MZ is, the larger the relaxation. With every recovery is relatively high. This is because expo-
small flip angle excitation, the MZ shrinks to the nential relaxation implies that a constant frac-
point at which the amount of MZ recovered tion of the remaining difference is recovered in a
because of longitudinal relaxation equals the given time increment. This situation is likely to
278 PART V  Pulse Sequences

t=0 t=1 t=2 t=3 t=4

BR BR BR BR BR

ω ω ω ω
MXY

t=5 t=6 t=7 t=8 t=9

BR BR BR BR BR

ω ω ω ω
MXY

B
FIGURE 18-9  A, In the presence of a gradient magnetic field, dephasing of the free induction decay increases more
rapidly and results in faster reduction in MXY. B, Reversing the gradient magnetic field focuses the spins to a weaker signal
intensity.

occur for the steady state of a large flip angle flip angle, only the component of the MZ that is
excitation or with short TRs. A decrease in lon- projected onto the XY plane produces a signal.
gitudinal magnetization caused by the large flip Consider the vector diagram in Figure 18-15.
angle excitation can only reach equilibrium with After an RF pulse of 20°, the MZ decreases only
an appropriate recovery during the time between 6% (Cos 20° = 0.94), whereas transverse mag-
excitations (TR). netization, MXY, which is the source of the MR
If the longitudinal magnetization is close to signal, increases to 34% of M0 (Sin 20° = 0.34).
equilibrium, the absolute recovery is relatively Increasing the flip angle of the α excitation RF
low. This situation often occurs with the steady pulse removes the MZ further from equilibrium
state of a low flip angle excitation or with rela- while increasing MXY. Table 18-1 shows these
tively long TRs. changing values for various flip angles. With flip
angles less than 60°, more than half of the lon-
gitudinal magnetization remains.
Low Flip Angle Assume that MXY is dephased or spoiled before
For NMR applications with a short TR, a 90° each α pulse in a steady state echo train. For
excitation may not necessarily provide the stron- extremely short TRs, low flip angles result in
gest magnetic resonance (MR) signal. For a low greater signal intensity than those obtained after
 CHAPTER 18  Steady State Gradient Echo Imaging 279

90° 180°

RFt

T2
T2*
RFs

SE
A

α
RFt

T2*

RFs

GRE
B
FIGURE 18-10  A, The spin echo (SE) is T2 dependent. B, The gradient echo (GRE) is T2* dependent.

the 90° RF pulse of SE with the same short TR

(Figure 18-16). At the TR, where MZ for an α
pulse crosses that for a 90° RF pulse, MZ is
almost at equilibrium for the α pulse while it is
still increasing for the 90° RF pulse.

Ernst Angle
Depending on the longitudinal relaxation time of
tissue (T1), the most intense MR signal in the
shortest TR time is produced with a specific flip
angle, the Ernst angle, which usually is less than
90°. For a given T1 relaxation time, the MR
signal per unit time might be higher with the use
of a flip angle lower than 90° (Figure 18-17).
Exciting with an RF pulse set to the Ernst
angle will produce the maximum signal that can
be achieved at a shorter TR. For longer TRs, the
FIGURE 18-11  Differences in magnetic susceptibility
among tissues at the base of the skull result in the  use of a low flip angle turns into a disadvantage
dephasing effects shown. (Courtesy Susan Weathers, because the available MZ is underused. For a
Houston, TX.) given TR and for a specific T1 relaxation time,
280 PART V  Pulse Sequences

α α

RFt

RFs

Stimulated echo
FIGURE 18-12  A stimulated echo is produced by refocusing the free induction decay.

α α α α α α
RFt

A B A A+B A+B+C B+C+ D

C D E F
RFs

FIGURE 18-13  Once the steady state is established, stimulated echoes are refocused from several preceding free induc-
tion decays. The result is strongly influenced by spin-spin relaxation.

MZ MZ MZ

TR TR TR

M0 M0 M0
MZ
MZ
MZ

MXY MXY MXY

FIGURE 18-14  A steady state is established.
 CHAPTER 18  Steady State Gradient Echo Imaging 281

M0
TABLE 18-1 Relative Longtitudinal and
0.94 M0 Transverse Magnetization
as a Function of Flip Angle
20° Immediately after
Radiofrequency Pulse

Flip Angle
(Degrees) Percent MZ Percent MXY
0.34 M0
0 100 0
FIGURE 18-15  This vector diagram shows how a small 10 98 17
α pulse can produce relatively large MXY. 20 94 34
40 77 64
60 50 87
90 0 100

90°

60°

30°
Relative
signal
10°

TR
FIGURE 18-16  For tissue with a given T1 relaxation time, signal intensity is higher with a flip angle that is less than 90°
and has a short repetition time (TR).

Long T1
TR

500 ms
Relative
signal
200 ms

Ernst 100 ms
angles

0 30 60 90
Flip angle (degrees)
FIGURE 18-17  The Ernst angle is the flip angle that produces the strongest signal at a given repetition time (TR).
282 PART V  Pulse Sequences

Ernst
α too low α too high angle

TR
1000 ms
T1 = 800 ms

Relative
signal

500 ms

200 ms
50 ms

0 30 60 90
Flip angle (degrees)
FIGURE 18-18  The Ernst angle is a function of repetition time (TR) and T1.

the Ernst angle is always the optimal flip angle the MZ and its recovered portion become involved
in which a maximum signal can be achieved in the next measurement.
(Figure 18-18). There is another group of GRE pulse sequences
that operates with a steady state for the MZ and
The shorter the TR, the smaller the flip angle the MXY. The first example of this type of GRE
necessary to achieve maximum signal intensity. sequence was called fast imaging with steady
state precession (FISP). Refocused GRE pulse
The optimum flip angle is a compromise sequences, such as FISP, aim to establish not only
between projecting enough magnetization onto a steady state in the longitudinal direction but
the XY plane to induce sufficient signal and also a steady state within the transverse plane
leaving enough in the longitudinal direction to (Figure 18-20).
use for the next excitation. This approach pro- Rather than spoiling the transverse magnetiza-
vides a different contrast than that observed in tion that has been dephased for the purpose of
SE imaging. spatial encoding, the magnetization is rephased
after data acquisition. This is accomplished by
designing the gradient waveforms so that the
SPOILED GRE VERSUS areas of positive and negative amplitude cancel
REFOCUSED GRE and the waveform is said to be “balanced.”
Spoiled GRE pulse sequences, such as FLASH Because this is repeated for each measured line
can use gradient magnetic field spoiling to destroy of k-space, it is expected that a steady state also
the phase coherence of MXY after the GRE signal develops in the transverse plane.
has been digitized (Figure 18-19). This approach
is similar to the crusher gradient technique used A pulse sequence with a gradient balanced in all
in SE imaging to avoid stimulated echo-related three directions is referred to as a balanced
artifacts. Another approach to eliminating the steady state free precession or bSSFP.
phase coherence of MXY is to change the phase
of each successive RF pulse for each measured There are different degrees of refocusing GRE
line of k-space, which is known as RF spoiling. pulse sequences, each of which produces slightly
Either of these spoiling methods ensure that only different forms of tissue contrast. In Chapter 17
 CHAPTER 18  Steady State Gradient Echo Imaging 283

RFt

Spoiler

GSS

RFs

Gφ

GR

FIGURE 18-19  A gradient magnetic field spoiler destroys the phase coherence of MXY.

α
RFt

BSS

RFs

rewinder

Bφ

BR

FIGURE 18-20  Some GRE sequences, such as FISP, establish the steady state in both MZ and MXY.

we noted that with three or more RF pulses maximize image SNR by re-encoding and reusing
applied in a time period less than T2*, three signals from the previous measured line of
types of signals can be generated; the FOD, the k-space.
SE, and the stimulated echo (STE). In Chapter 17 In practice, there are three types of refocused
we showed how in FSE imaging crusher gradi- GRE sequences. First, there is the basic refocused
ents are used to eliminate the FID and STE signals GRE sequence, for example FISP, which acquires
in order to detect a train of pure SE signals. In a signal that is composed of both an FID and
refocused GRE imaging, two or more of these STE signals. Second, there is the contrast-
signals can be detected together in order to enhanced GRE sequence, such as PSIF, which
284 PART V  Pulse Sequences

acquires a signal that is composed of both an either tissue. Also, because the steady state in the
STE and SEs. Third, there is the balanced steady transverse plane, MXY, must be developed and
state free precession sequence (bSSFP), like True maintained over several repetitions, T2* dephas-
FISP, which acquires a signal that is composed of ing tends to destroy this coherence.
FID, STE, and SE signals. Consequently, the shorter the TR, the better
A difference in image appearance between the image contrast. In general, refocused GRE
spoiled GRE and refocused GRE occurs only if pulse sequences produce poorer contrast but
certain conditions are fulfilled. First, only tissues better SNR than spoiled GRE pulse sequences.
with a long T2*, relative to TR, show an At flip angles, α > 20°, spoiled GRE images
enhanced signal. If the ratio T2*/TR is less than exhibit substantial Tl-weighting and fluids, such
about 10, there is too much dephasing to build as blood, appear dark. Refocused GRE images
up a steady state of the transverse component. appear to have less contrast but fluids in these
This means that most GRE pulse sequences have images are bright.
TR < 10 ms, often less than 5 ms. Second, the Finally, the contribution to the transverse
transverse component must be nurtured accord- steady state should appear at the same phase for
ing to the projection of the longitudinal compo- a constructive interference. Any motion of the
nent. Standard manufacturer names for refocused patient that changes the phase history does not
GRE pulse sequences include GRASS, FFE, FISP, provide the necessary phase coherence, and the
and rephased SARGE. transverse steady state is not established.
The larger the flip angle, the bigger the con-
tribution to the transverse steady state. It is
CONTRAST-ENHANCED
important to keep in mind that there is a distinc-
TECHNIQUES
tion to be made between maximizing the SNR
and maximizing the image contrast. Although Once the steady state is reached, which can take
the maximum SNR per unit time is achieved by a number of α pulses, signal intensity occurs
applying the formula for the Ernst angle, because of both the FID and the SE (Figure
described above, the optimum T1-weighted con- 18-21). A GRE obtained from this signal at echo
trast between two tissues is actually obtained at time (TE) is generated with the summed signal of
a α flip angle that exceeds the Ernst angle for both the FID and the SE. Such a GRE mixes

TR
α
RFt

FID + SE
FID

RFs

SE
TE

BR

FIGURE 18-21  In the steady state, signal intensity occurs because of both the free induction decay (FID) and the spin
echo (SE).
 CHAPTER 18  Steady State Gradient Echo Imaging 285

TR
α α
RFt

FID + SE

FID

RFs

SE
TE

BR

FIGURE 18-22  In the steady state, suppressing the spin echo (SE) component of the signal results in T1-weighted
contrast.

contrast dependent on proton density (PD), T1, various MRI system manufacturers. The data are
and T2 as determined by the values of TR, first acquired, then comes the phase encoding,
TE, and α. and finally the excitation (Figure 18-24).
Although PSIF looks like a violation of causal-
Contrast manipulation in GRE imaging is accom- ity, it is not. The signal collected with a PSIF
plished by precisely timing the read gradient technique has been prepared in the previous exci-
magnetic field, GR. tation. Because the echo is created in the previous
excitation, the effective echo time is approxi-
If the phase of the α pulse reverses or cycles mately twice the TR. That explains the relatively
properly, the SE component of the signal can be heavy T2-weighted results, compared to other
suppressed (Figure 18-22). The GRE formed GRE pulse sequences.
under this condition only reflects the FID char- GRE pulse sequences producing SE-type
acter and exhibits T1-weighted contrast, as contrast are rarely used in clinical practice since
described above. they cannot compete with FSE in terms of SNR
If the refocusing gradient is energized before and T2-weighted contrast. However, the method
the α pulse (Figure 18-23), the result is the same may find niche applications. For example, this
as removing the FID. With this the GRE is formed technique can be used to study the inner ear
that has an SE character and results in a more (Figure 18-25).
T2-weighted image. Adding the FISP echo and the PSIF echo
An example of this second type of refocused increases the T2 weighting of the basic FISP
GRE is precession of steady state imaging fast image. This approach allows a better delineation
(PSIF), which is an SSFP sequence that produces of fluid, fat, and cartilage (Figure 18-26), making
an RF SE, rather than a refocused FID. The it promising in orthopedic work. This is called
name, PSIF, shows that it is a time-reversed FISP. double echo steady state (DESS).
This type of pulse sequence is also identified as The third approach for producing refocused
SSFP, CE-FFE-T2 and time-reversed SARGE for GRE images aims at producing a completely
286 PART V  Pulse Sequences

TR
α α

RFt

FID + SE

FID

RFs

SE

TE

BR

FIGURE 18-23  In the steady state, suppressing the free induction decay (FID) component of the signal results in
T2-weighted contrast.

α α α
RFt

BSS

RFs

Bφ

BR

FIGURE 18-24  The PSIF pulse sequence.

balanced GRE pulse sequence. This approach is equal to zero from one RF pulse to the next
known generically as balanced steady state free (Figure 18-27). This preserves as much signal as
precession (bSSFP), but also goes under the trade possible for each phase-encoding k-space line of
names of True FISP, balanced FFE, FIESTA, data acquired. If the α RF pulses are phase-
BASG, and True SSFP. inverted on alternating acquisitions and the bal-
The bSSFP sequence uses symmetric gradient anced gradient configuration is applied with a
waveforms to generate a net gradient moment short TR (i.e., TR << T2 < T1), then the balanced
 CHAPTER 18  Steady State Gradient Echo Imaging 287

GRE signal will be both proportional to T2 and is acquired at TE = TR/2, dephasing effects (due
inversely proportional to T1. This is used to to chemical shifts or off-resonance) are nearly
produce a hyperintense signal from fluids and is completely refocused.
often applied to cardiac imaging and MRA pulse The biggest issue with balanced GRE pulse
sequences. sequences is that off-resonance signals drop off
Balanced SSFP produces mixed image contrast severely. The signal amplitude in bSSFP sequences
that is proportional to T2/T1. This results in very depends on all the spins resonating at the same
high signals whenever T1 is similar to T2 (liquids, frequency. The balanced gradients aim at pro-
fat) and low signals for long T1 and short T2 ducing this effect, but if there are significant B0
tissues, such as muscle. If a balanced GRE signal magnetic field inhomogeneities the signal may

FIGURE 18-25  A three-dimensional PSIF image of the FIGURE 18-26  A three-dimensional DESS image of the
inner ear (TR = 17 ms, effective TE = 30 ms). (Courtesy knee showing destruction of cartilage (TR = 28 ms, effec-
Susan Weathers, Houston, TX.) tive TE = 52 ms). (Courtesy Tom Hedrick, Houston, TX.)

rf
TX
Field
echo
RX
Slice select Digitizer on
Gsl
Rephasing Read out
Dephasing
Gro
Phase Phase
encode rewinder
Gpe

Also called FAST, ROAST

FIGURE 18-27  The balanced steady state free precession pulse sequence (True FISP).
288 PART V  Pulse Sequences

B
FIGURE 18-28  This cardiac True FISP image exhibits dark bands near the periphery, which are due to off-resonance
effects resulting from magnetic field inhomogeneities. As long as these bands are kept away from the heart, images will
still be useful.

drop close to zero, producing strong, dark Whenever a GRE application uses a flip angle
banding artifacts in bSSFP images (Figure 18-28). between 30° and 70°, it is probably an Ernst
For this reason it is important to make the B0 angle approach. The Ernst angle is also deter-
magnetic field as homogeneous as possible, using mined for time-of-flight (TOF) angiography in
localized auto-shimming and also to minimize which both relaxation of blood and tissue
TR where possible for bSSFP. replacement need to be considered. The same
On the basis of the refocused GRE method, criterion, maximizing signal strength by choos-
new clinical applications have become well estab- ing the optimum flip angle, is also used in ultra-
lished. Some examples include applications in fast GRE imaging.
orthopedics (Figure 18-29) and dynamic studies The absolute relaxation is lower when most
of the heart with the use of breathhold imaging of the net magnetization lies along the Z-axis.
techniques (Figure 18-30). Although that relaxation is a function of the
 CHAPTER 18  Steady State Gradient Echo Imaging 289

longitudinal relaxation time, T1, for that tissue,

the differences in absolute relaxation of net mag-
netization are small for even great differences in
T1. If only a small part of the available net mag-
netization is projected onto the XY plane, while
most magnetization remains in the longitudinal
direction, then only a slight disturbance from
equilibrium occurs and the absolute relaxation is
small.
Such a recovery rate is almost independent of
the T1. Using an extremely low flip angle (i.e.,
5° to 15°) establishes a situation in which the
MR signal strength is almost constant, even with
shorter TR times and tissues with significant dif-
FIGURE 18-29  A two-dimensional example of a meniscal ferences in T1 relaxation times.
tear (TR = 560 ms, TE = 10 ms, α = 60°, 7 min). (Courtesy
Tom Hedrick, Houston, TX.)
CHALLENGE QUESTIONS
1. What is the principal advantage to GRE
imaging?
2. What are the general characteristics of Ernst
angle imaging?
3. What mechanism is used to rephase spins
during GRE imaging?
4. What characterizes SSFP imaging?
5. Why is the read gradient magnetic field, GR,
bipolar in GRE imaging?
6. Graphically illustrate the difference between
T2 and T2*.
7. What principally determines the value of
T2*?
8. What is the difference between spoiled GRE
and refocused GRE?
FIGURE 18-30  A two-dimensional image in a dynamic
study of a long axis view of the heart showing the aortic 9. What is the Ernst angle?
outflow tract and the mitral valve (TR = 40 ms, TE = 7 ms, 10. How does signal intensity vary with flip
α = 30°, 6 min). (Courtesy Tom Hedrick, Houston, TX.) angle from 5° to 90°?
 CHAPTER

19 
Hybrid Fast Imaging
Techniques

OBJECTIVES
At the completion of this chapter, the student • Describe sequential filling of k-space and its
should be able to do the following: application.
• Identify the pulse sequence characteristic of • Define isotropic imaging and the application
turbo imaging. of MPRAGE technique.
• Explain the use of a preparation pulse in fast • Draw a GRASE pulse sequence.
imaging.

OUTLINE
Turbo Imaging
Gradient and Spin Echo
Imaging

KEY TERMS
Echo train Inversion pulse Turbo
Filter Segmentation

Speeding up image acquisition by reducing the technique. Recall that in IR SE, the inversion of
repetition time (TR) and flip angle, α, causes a spins occurs before each sampled line of k-space.
loss of contrast in gradient echo (GRE) images.
Those parameters, which are usually available to
TURBO IMAGING
improve contrast, are constrained by speed and
signal-to-noise requirements. These methods often have names with a turbo,
Improved T1-weighted (T1W) contrast can snapshot, insta, or hyper prefix. GREs are
be obtained in a fashion similar to that used in acquired after several α pulses prepared by an
spin echo (SE) imaging with spin preparation inversion pulse (Figure 19-1). For each line of
strategies, such as the inversion recovery (IR) k-space, a different spatial frequency in the

290
 CHAPTER 19  Hybrid Fast Imaging Techniques 291

α α α
RFt

BSS

GRE #1 GRE #2 GRE #128

RFs

Bφ

BR

FIGURE 19-1  Data acquisition for turboFLASH imaging techniques.

direction of phase encoding is sampled with each The common approach to turbo imaging is
phase-encoding step. a 180° inversion pulse at the beginning of the
sequence to generate T1W contrast. Other spin
Low spatial frequencies, which contain the preparation methods have been introduced to
patient’s coarse structure, follow weak phase- provide T2-weighted (T2W) contrast or flow-
encoding gradients. diffusion weighting. With this technique, the
data are acquired while the tissue relaxes.
Low spatial frequencies are more important Each line of k-space deals with a different
to image contrast than high frequencies, which amount of longitudinal magnetization (MZ). In
primarily contribute to edge enhancement and other words, each line has a slightly different
spatial resolution. Understanding this spatial fre- T1W. The lines of k-space acquired with low-
quency weighting is the key to understanding amplitude phase-encoding gradients that sample
both the image contrast behaviors of faster the lower spatial frequencies are important for
imaging and the way that a static image can be image contrast.
obtained from a moving object. For certain clinical applications, these turbo
techniques challenge the established GRE tech-
High spatial frequencies, which contain the nique. Although the GRE technique provides an
patient’s fine structure, follow strong phase- image within 10 seconds, combining GRE with
encoding gradients. IR, while using even the shortest TI mixing times,
would make the GRE sequence take an unaccept-
TurboFLASH, snapshotFLASH, snapshot- ably long time. The problem then is how to use
GRASS, and others are all similar imaging spin preparation schemes in a manner that pre-
techniques that were introduced at different serves the short overall image acquisition times
times by different groups who worked with dif- of GRE.
ferent equipment. In this discussion, these tech- One-second images bring to mind new clinical
niques are identified as “turbo” because that is applications; however, the simultaneous mul-
the term most adopted. tislice GRE technique is almost as fast as the
292 PART V  Pulse Sequences

A B
FIGURE 19-2  A, Breathhold liver image gradient echo. B, Breathhold liver image with turboFLASH.

sequential multislice turbo approach in abdomi- similar to FSE in which the TE was different
nal imaging within a breathhold. Nevertheless, from spin echo to spin echo, except in this case
because the turbo technique results in an image it is the TI that is different from gradient echo to
one slice at a time, there is a noticeable reduction gradient echo.
in motion artifacts.
When compared with GRE, turbo imaging has The order in which k-space is filled during turbo
additional degrees of freedom to control the con- imaging adjusts image contrast.
trast within the image. In Figure 19-2 a breath-
hold liver study acquired with a GRE technique With turbo imaging, this steady state condi-
is compared with one acquired with the turbo- tion is no longer fulfilled. The k-space–ordering
FLASH technique. scheme becomes an adjustable parameter to
These turbo imaging techniques significantly achieve the desired contrast. Figure 19-3, A,
increase the diagnostic capabilities of magnetic shows that k-space is being filled sequentially
resonance imaging (MRI) in applications like the from the smallest phase-encoding gradient to the
imaging of the first pass of contrast agents, such largest. In Figure 19-3, B, the filling of k-space
as Gd-DTPA (gadolinium-diethylene-triamine- is ordered by particular selection of the ampli-
pentaacetic acid). Studying the temporal course tude of the phase-encoding gradient. This tool is
of Gd-DTPA distribution within liver and breast valuable because the course for the relaxation of
lesions or for brain perfusion studies results in MZ is altered depending on the flip angle used
a better temporal resolution and sometimes a throughout the signal acquisition.
better signal-to-noise ratio (SNR) than GRE Another degree of freedom for adjusting con-
imaging. trast occurs by changing the flip angle of the
When the available magnetization changes radiofrequency (RF) excitation pulse from one
over time, as in the turbo methods, another Fourier line to another. An RF pulse of a given
imaging parameter becomes an issue: especially angle excites the same fraction of the Z-axis
the scheme of filling k-space. The order in which magnetization regardless of the degree of relax-
k-space was filled in SE or GRE imaging was ation. Thus the early Fourier lines would be
unimportant because the acquisition began after measured from stronger transverse magnetiza-
reaching a steady state. Each line of k-space dealt tion (MXY) than later lines when the flip angle
with the same amount of MZ. Turbo GRE is increases for the later Fourier lines. This strategy
 CHAPTER 19  Hybrid Fast Imaging Techniques 293

α 180° α
RFt RFt

BSS BSS

RFs RFs

Bφ Bφ

BR BR
A B
FIGURE 19-3  A, The k-space is normally filled sequentially from the lowest line up. B, The filling of k-spaces is select-
able in turbo imaging to adjust contrast.

180°
α α α
RFt

BSS

RFs

Bφ

BR

FIGURE 19-4  The concept of segmentation of Fourier lines.

drives MZ (and subsequently MXY) to a steady TR. This would result in an imaging time of
state more quickly. 640 ms. Such an imaging time is probably still
It is possible to use a larger fraction of the too long to freeze heart motion, but this defi-
Z-axis magnetization, thereby keeping the mag- ciency can be partially corrected by segmentation
nitude of MXY constant from line to line of (Figure 19-4).
k-space. This makes it possible to reduce the In SE and GRE imaging, this problem is solved
blurring effects observed for large differences in by triggering the acquisition for each Fourier line
signal response from one Fourier line to another. to correspond with a fixed time in the cardiac
Consider a turbo technique applied to the cycle. In segmented GRE imaging, a certain
heart with 64 phase-encoding steps and a 10-ms number of Fourier lines are acquired (e.g., seven
294 PART V  Pulse Sequences

A B
FIGURE 19-5  Short axis views of the heart acquired with varying degrees of segmentation.

lines) so that a 133 × 256 matrix can be filled difference in contrast between turbo and
within 19 heartbeats. MPRAGE techniques.
Such a technique allows dynamic imaging of Using a varying flip angle reduces artifacts
the heart with breathholding to reduce respira- further. Spatial frequency filtering causes reduc-
tory motion. Figure 19-5, A, is a segmented fast tion by shaping the signal’s approach to steady
low-angle shot (FLASH) with five lines per state. Acquiring data while MZ relaxes applies a
segment acquired in 1 min, 12 s. Figure 19-5, B, spatial frequency filter to the ideal spatial fre-
was obtained in 16 heartbeats (one breathhold) quency spectrum.
with nine Fourier lines per segment.
Mugler at the University of Virginia is Spatial frequency filtering changes apparent
credited with introducing the three-dimensional spatial resolution and contrast resolution.
(3D) version of turbo techniques, called the
magnetization prepared rapid gradient echo The term filter in this context comes from elec-
(MPRAGE) technique. In two-dimensional (2D) trical engineering in which the relative strengths
turbo imaging, the net magnetization is prepared of high and low spatial frequencies are modified.
before acquiring all the Fourier lines. To simply When high spatial frequencies are amplified, edge
apply this approach to a 3D acquisition scheme sharpness increases. When low spatial frequen-
is not advised because more excitations alter the cies are amplified, blurring increases and appar-
course of the relaxation and destroy the initial ent contrast resolution improves.
preparation. The same thing happens when measuring
The solution is the use of turbo techniques for the high or low spatial frequencies. When the
all in-depth phase-encoding steps. A recovery Z-axis magnetization is nearly relaxed, an
period is allowed before proceeding to the next emphasis of the high or low spatial frequencies
in-plane, phase-encoding step, which begins produces an edge sharpness or blurring, respec-
again with the inversion to prepare MZ (Figure tively. As a result, the original object is inaccu-
19-6). The amount of MZ is the same for each rately represented.
step for in-plane phase encoding. For the depth- This spatial frequency filter is likely to produce
encoding loop, the signal acquisition is done image artifacts that appear as blurring, ringing,
along the relaxation path. This accounts for the and edge enhancement. With variation of the
 CHAPTER 19  Hybrid Fast Imaging Techniques 295

α α
RFt

BSS

RFs
Bφ Bφ

BR BR

Bφ

BR

FIGURE 19-6  The MPRAGE pulse sequence.

in which the entire head is imaged in approxi-

mately 7 minutes with an isotropic resolution of
1.25 mm.

GRADIENT AND SPIN

ECHO IMAGING
Feinberg and Oshio introduced the combination
of fast spin echo (FSE) and GRE, which they
called gradient and spin echo (GRASE) imaging.
Rather than only one echo being sampled between
180° refocusing pulses, three GREs are acquired
(Figure 19-8). The principal advantage of this
technique is a significantly shorter echo train,
FIGURE 19-7  Image acquired with MPRAGE with isomet-
ric resolution of 1.25 mm and no gaps.
which leads to more slices per TR and a lower
RF power deposition.
Because the J-coupling pattern remains intact,
excitation angle from one depth-encoding step this sequence also provides a fat intensity similar
to the next, it is possible to achieve a signal to SE techniques rather than FSE methods. The
course where this spatial frequency filter effect is GRASE technique has a higher sensitivity to
eliminated, resulting in fewer such artifacts. magnetic susceptibility, which is typical for GRE
Figure 19-7 represents a T1W study of the head imaging techniques.
296 PART V  Pulse Sequences

90° 180° 180° 180° 180°

RFt

SE1 SE2 SE3 SE4

RFs

SE2

BR

GRE1 GRE3
GRE2
RFs

FIGURE 19-8  The pulse sequence for GRASE imaging.

 CHAPTER 19  Hybrid Fast Imaging Techniques 297

CHALLENGE QUESTIONS
1. What is turbo imaging?
2. What does GRASE identify?
3. How does turbo imaging differ from SE or
GRE imaging in regard to the filling of
k-space?
4. Which of the following pulse sequences
results in more patient RF exposure: SE, IR,
GRE, FSE, or GRASE?
5. Why is ordering of k-space so important for
turbo imaging?
6. What is meant by the term high pass filter?
7. What is meant by segmental imaging in
cardiac studies?
8. Diagram the RF pulse sequence and signal
acquisition sequence for a turboFLASH
imaging technique.
9. What does the symbol Gф with the step indi-
cation mean?
10. What is isotropic imaging? "I assure you … your magnetic personality
won't affect our MRI."
 CHAPTER

20 
Echo-Planar Imaging

OBJECTIVES
At the completion of this chapter, the student • Identify the enhanced hardware requirements
should be able to do the following: for echo-planar imaging.
• Diagram an echo-planar imaging pulse • Report on several clinical situations that are
sequence. particularly suited to echo-planar imaging.
• Describe the multiple ways that k-space is
filled during echo-planar imaging.

OUTLINE
Echo-Planar Imaging
Hardware Requirements

KEY TERMS
Amplitude gradient switching Blood oxygen level dependent
Blipped echo planar K-space trajectory

Echo-planar imaging (EPI) is a method for require separate spin excitation as in spin echo
extremely fast formation of the magnetic reso- (SE) or gradient echo (GRE) imaging.
nance (MR) image. Some of its commercial The entirety of k-space can be sampled and
implementations have trade names that imply filled by measuring either the envelope of
instantaneous image acquisition. Although that gradient-refocused free induction decay (FID) or
may be a slight exaggeration, imaging times of that of an SE. In either case, the entire image is
as little as 50 ms to produce images of moderate acquired following a single RF excitation and
quality are realistic. within one repetition time (TR).

The fundamental idea of EPI is to fill k-space with

ECHO-PLANAR IMAGING the magnetization produced by a single RF pulse.
In 1977 Mansfield first introduced the concept
of EPI. The distinguishing characteristic of EPI is SE and GRE techniques measure only a
the filling of k-space after a single radiofrequency portion of k-space in each TR interval. That
(RF) excitation. Each line of k-space does not portion is typically one Fourier line of k-space.

298
 CHAPTER 20  Echo-Planar Imaging 299

180°
90°

RFt

BSS

SE envelope

FID

RFs

GRE

Bφ

BR
FIGURE 20-1  The basic pulse sequence for echo-planar imaging.

Thus T1 is a consideration because it is necessary generating a GRE each time. In its original form,
to allow longitudinal magnetization to recover demonstrated by Mansfield, the phase encoding
before the next portion of k-space can be is done with a small constant gradient through-
measured. out the acquisition. This way, each echo has a
EPI avoids this problem by measuring all of different phase encoding and can be used to fill
k-space in one pass. However, because this one k-space. Acquisition times are about 10 times
pass may be rather long, spin-spin relaxation faster than those for the GRE imaging techniques
(T2) during signal sampling becomes an issue. described in Chapter 18.
The raw data measurements must be made The train of echoes obtained in EPI is pro-
rapidly to measure all of k-space before the duced by switching the gradient magnetic field
transverse magnetization MXY is substantially and sampling the signal in the envelope of the
altered in magnitude by T2* relaxation. SE. EPI GREs are produced with modifications
Thus the magnetic resonance imaging (MRI) of the scheme shown in Figure 20-1.
system must be capable of extremely fast and The original echo-planar method proposed by
high-amplitude gradient switching and rapid Mansfield was complicated by limitations of his
data acquisition. Manufacturers have solved hardware. A more recent version, blipped echo
these engineering and hardware requirements. planar, has been universally adopted in commer-
A single RF excitation α pulse is followed by cial MRI systems.
a 180° RF refocusing pulse. The MR signal read Figure 20-2 shows the pulse sequence diagram
phase contains a train of GREs produced by of a blipped echo-planar image. Figure 20-3
rapidly switching the read gradient magnetic shows the path through k-space that is traversed
field (GR) (Figure 20-1). in sampling the GRE signals.
After a single excitation, the read gradient There is only one RF pulse. The phase-
is switched from one polarity to another, encoding gradient waveform consists of a string
300 PART V  Pulse Sequences

RFt

BSS

RFS

Bφ

BR
FIGURE 20-2  A blipped echo-planar pulse sequence.

of short, relatively weak “blips” (hence, the The k-space trajectory that is created by the
name of the sequence). Between each blip, the blipped echo-planar sequence is designed to be
frequency-encoding gradient waveform is turned suitable for the Fourier transform to process the
on rapidly to a maximum value and then turned data. There is still one small problem though:
off just before the next blip. Signals are acquired alternate lines scanned in the reverse direction
between the blips. in the kx direction. This means that for the
The frequency-encoding read gradient mag- odd lines the signal is decaying from left to
netic field, GR, alternates between positive and right, while for the even lines the signal is
negative amplitude. This is the reason the path decaying from right to left. Thus to minimize
in the k-space goes back and forth. The phase- artifacts, alternate lines in the raw data matrix
encoding gradient pulse moves the path to must be reversed before the Fourier transform
the left edge of the k-space at the same time the is applied.
frequency-encoding gradient pulse moves the Figure 20-4 shows images of the same slice
path to the bottom. from the same patient. The SE image (Figure
The acquisition of data starts during the GR 20-4, A) was acquired in 7 min, 8 s. The echo-
pulse, which moves the path from left to right. planar image (Figure 20-4, B) was acquired in
The blip pulse shifts the path up one row. Then 120 ms. There is still only one RF pulse used for
another GR pulse, which is of the opposite ampli- excitation with EPI.
tude, moves the path from right to left. The The transverse magnetization is prepared by
phase-encoding gradient blips shift the path to the 90° and 180° RF pulses, and then the regular
different rows, and the frequency-encoding gra- blipped echo-planar gradient waveforms follow.
dient pulses sweep the path along a row. T1-weighted (T1W) sequences can be constructed
 CHAPTER 20  Echo-Planar Imaging 301

KY (phase) KY (phase)

KX (frequency)

2DFT EPI
FIGURE 20-3  The path through the k-space traversed in blipped echo-planar imaging.

A B
FIGURE 20-4  T2-weighted images acquired by (A) spin echo in 7 min, 8 s and (B) echo-planar imaging in 120 ms.
302 PART V  Pulse Sequences

in a similar fashion by placing RF pulses that EPI requires special MRI hardware and
implement a partial saturation or an inversion data-handling capacity. However, a number of
recovery (IR) contrast weighting in front of the methods similar to echo-planar that do not
echo-planar acquisition. require such exotic hardware have been pro-
posed and investigated. The gradient and spin
echo (GRASE) method mentioned in Chapter 19
HARDWARE REQUIREMENTS is an example of a hybrid of fast spin echo (FSE)
The hardware requirements of EPI include and EPI. Between each pair of 180° RF pulses,
switching the frequency-encoding gradient and not just one SE but several GREs are formed.
receiving the MR signals rapidly and nearly con- Those GREs are the same as the echoes generated
tinuously. These hardware requirements were the by the echo-planar approach.
original impetus for the development of shielded There are a number of methods for filling
gradient coils; the eddy currents induced in k-space in patterns such as spirals, square spirals,
unshielded gradient coils destroy the EPI signal, and rosettes. These methods map the entire
even if they are sufficiently small to allow routine k-space with the magnetization from a single RF
SE and GRE imaging. pulse, although some of the implementations of
Exceedingly fast, high-intensity gradient mag- these ideas segment the acquisition and use a few
netic fields are required for EPI. Rise and fall RF pulses to cover all of k-space.
times of approximately 100 µs are required. Gra- However, these approaches are limited because
dient magnetic fields exceeding 25 mT/m are they tend to oversample the center of k-space
common. and undersample the peripheral regions, which
The attractions of EPI have encouraged engi- results in poorer spatial resolution. Also, since
neering to overcome these technical obstacles. these data obtained using these methods are not
The use of a single excitation RF pulse means sampled on a rectilinear grid, the data must be
that the echo-planar image reflects a brief inter- interpolated to the proper positions in k-space
val of time. before the Fourier transform is performed, taking
up many processing cycles.
Echo-planar images can be obtained at radio- Image contrast can be readily manipulated in
graphic speed, in as little as 50 ms. EPI to accentuate tissue differences of proton
density (PD), T1, or T2. Spin preparation for fat
SE and GRE techniques must image dynamic or water suppression, chemical shift imaging,
processes by synchronizing to the physiologic T1W with an inversion pulse, and flow encoding
cycle, as in cardiac imaging. It is assumed that are possible.
the variations from cycle to cycle are minimal to
measure each line of k-space at the same phase The principal advantage to EPI is the ability to
of a physiologic cycle. freeze motion.
EPI excites the spins once and then rapidly
acquires all the data needed for the image. In With EPI image acquisition times of 50 to
its faster implementations, EPI has temporal 100 ms, even cardiac motion, up to 10 cm/s, is
resolution equivalent to ultrasound and electron reduced as in a radiographic image. Indeed, EPI
beam computed tomography (EBCT). Thus it represents cineradiography-like MRI. Cardiac
is possible to image patients with cardiac imaging was the application that motivated
arrhythmias and to look at fast processes in the Mansfield to develop EPI. However, in current
brain, such as water diffusion, oxygen use, and practice, cardiac imaging more often employs the
contrast agent dynamics. EPI is also used for segmented GRE methods, mentioned in Chapter
dynamic imaging of the viscera and the human 19, and the primary applications for EPI are in
fetus. neuroimaging.
 CHAPTER 20  Echo-Planar Imaging 303

A concern about the use of EPI has been the sequence plays out. If B0 homogeneity is poor,
physiologic stimulation threshold. A rapidly significant image distortion may be present,
changing magnetic field (dB/dt) may result in which can hinder attempts to fuse EPI images
nerve stimulation through magnetic induction, with other MRI image data or images produced
leading to muscle contraction. Very powerful using other image modalities.
gradient magnetic fields are necessary to produce Typical applications for EPI are first-pass
high-resolution images or ultrashort imaging tracing of contrast agents for perfusion imaging,
times. functional MR imaging (fMRI) using blood
The required rapid switching of these gradi- oxygen level dependent (BOLD) contrast, and
ents can produce magnetic field changes up to diffusion imaging of water. Single-slice EPI can,
100 T/s. This is above the reported neural stimu- in principle, be acquired in a manner that is
lation threshold of approximately 60 T/s. Ven- independent of TR. However, in practice EPI is
tricular fibrillation has been induced at 250 T/s. often used to follow physiologic changes over
Most currently developed EPI techniques are time. In this case up to 20 slices are scanned and
well below that stimulation threshold and repeatedly rescanned with TR intervals of any-
modern MRI scanners have safety systems built where from 4 to 20 seconds.
in which monitor the dB/dt to avoid muscle stim-
ulation during scanning.
CHALLENGE QUESTIONS
Various resonance offset artifacts appear in
EPI images. The FOV/2 ghost also called the 1. What is the fastest mode of MR image acqui-
“Nyquist ghost” is a ghost that is displaced sition and how fast is it?
exactly one-half the field-of-view. This is due to 2. What are the principal applications for EPI?
imperfections of the rephasing-dephasing cycle 3. Which of the following RF pulse sequences
of the rapidly switching frequency-encoding gra- deposits the least energy in tissue for a given
dient, including gradient instabilities, eddy cur- image: SE, IR, FSE, GRE, EPI, or GRASE?
rents, and poor B0 magnetic field uniformity. 4. Why is fat saturation technique often re-
Another problem for practical EPI is the quired for EPI?
chemical shift artifact. Unlike the case for SE and 5. How can an entire image be formed with a
GRE, the chemical shift artifact presents in the single SE signal?
phase-encoding direction and can be 20% to 6. What particular potential hazard is associ-
30% of the field of view, rather than just a few ated with EPI?
pixels. It is necessary to use some sort of tech- 7. Which of the following nuclear magnetic
nique, such as fat saturation, to minimize the resonance (NMR) parameters—PD, T1, or
intensity of the fat signal and avoid unacceptable T2—is changing during signal acquisition in
degradation of the echo-planar image. EPI?
Also magnetic susceptibility artifacts, which 8. Which prevails in EPI: PDW, T1W, or T2W?
are seen on other images, such as GRE, are most 9. What is a shielded gradient coil, and why is
severe in EPI images due to the long acquisition it important for EPI?
time (50 to 80 ms) that allows phase errors in 10. What principal hardware characteristics are
the magnetization to accumulate as the pulse required for EPI?
PART
VI
Applications
 CHAPTER

21 
Nuclear Magnetic
Resonance Spectroscopy

OBJECTIVES
At the completion of this chapter, the student • Name five metabolites with potential for
should be able to do the following: patient magnetic resonance spectroscopy
• Describe a frequency distribution and (MRS).
spectrum. • Define chemical shift.
• Explain the meaning of high-resolution • Discuss the use of the PPM (parts per million)
nuclear magnetic resonance (NMR) scale.
spectroscopy. • Identify J-coupling.

OUTLINE
Nuclear Species J-Coupling Sodium
Chemical Shift Medically Important Nuclei Fluorine
Magnetic Field Dependence Hydrogen Nitrogen
The PPM Scale Phosphorus
Signal Intensity Carbon

KEY TERMS
Downfield Frequency distribution Spectrum
Chemical shift Spectroscopy Upfield

This chapter has several goals: to explain basic medical companion magnetic resonance spec-
nuclear magnetic resonance (NMR) spectros- troscopy (MRS) is extensive.
copy, to get the reader excited about the use of Magnetic resonance imaging (MRI), the prin-
NMR in medicine, and to show that expert treat- cipal subject of this book, evolved from the
ment of NMR spectroscopy is a specialty topic. scientific application of NMR high-resolution
The treatment is thus illustrative, not compre- spectroscopy. Knowledge of NMR spectroscopy
hensive. The scientific literature on NMR and its is not essential for an understanding of MRI.

305
306 PART VI  Applications

Rather, it is an enhancement that should be in the FID results from the excitation of an NMR
the tool chest of any serious MRI radiologist or sample by a single RF pulse, a mathematical
technologist. process called the Fourier transform (FT) can be
This chapter deals with NMR spectroscopy in used to convert the abstract FID to a readily
simple terms. It will give the imaging physician interpreted spectrum. The NMR spectrum is pre-
and technologist a brief look into the history of sented as signal intensity versus frequency.
NMR and a similar glance at the future. It will Fourier transformation is used to analyze peri-
also give the chemical NMR spectroscopist an odic (sine or cosine) wave functions mathemati-
equally interesting view of the future. One prin- cally. Knowledge of the operation or underlying
cipal aim of MRS use in MRI is the performance mathematics of the FT is not required for the
of in vivo analysis of pathologic conditions, production of MR images or spectra. The reader
which increases the diagnostic value of MRI. can simply accept the FT as a black box that
The first concept to remember is the spectrum. produces acceptable results from FID data sets.
A spectrum, sometimes referred to as a frequency Consider an analogy to poker strategies
distribution, is a convenient graphic means of (Figure 21-4). One strategy is to make all moves
presenting specific frequency or wavelength- on the basis of experience, the ability to keep a
related material. Two spectra that should be poker face, and a good feel for the game. Another,
familiar are those associated with the emission more technical, strategy is to factor in the role of
of light from rare-earth radiographic intensifying probability in the construction of the hand and
screens and the absorption of that light in the to discard and draw more cards to improve the
emulsion of radiographic film (Figure 21-1). chances of making a better hand. At the end of
Absorptiometry is an analysis of the absorp- play, a player using the first strategy will prob-
tion of ultraviolet (UV), visual, and infrared ably admit that poker is fun but a little expen-
(IR) light transmission through a sample. If the sive. A more technical player may or may not
measurement is made at many different wave- enjoy the game as much but will probably have
lengths or frequencies, the resulting graph is extra money.
reported as a spectrum. For NMR spectroscopy
the X-axis defines the radiofrequencies (RFs)
NUCLEAR SPECIES
emitted by the patient, and the Y-axis is the
intensity of the signal produced by the patient at Shortly after the discovery of NMR, one princi-
each frequency. Representative UV absorption, ple became well known: each nuclide in the peri-
IR transmission, and NMR emission spectra are odic table has a unique resonance frequency at
shown in Figure 21-2. a given magnetic field strength. In the Larmor
The line widths seen in Figure 21-2 are differ- relationship the resonant frequency of any NMR
ent in these three spectra. Those of the first two signal depends only on the gyromagnetic ratio
methods are “broad,” whereas the NMR signals and the magnetic field strength. The value of
appear “sharp.” Indeed, NMR is widely regarded gyromagnetic ratio varies by a factor of more
as high-resolution spectroscopy. than 100 among nuclides.

MRS line widths can be exceedingly sharp. Each nuclide has its own gyromagnetic ratio.

An NMR spectrum can be obtained from any The gyromagnetic ratios of several nuclides
free induction decay (FID). The FID is the NMR of interest to MR in medicine are given in
signal emitted at the Larmor frequency from an Table 3-1. If a total NMR spectrum of the body
excited sample. The FID may be viewed as a plot could be produced, it would include many
of signal intensity versus time (Figure 21-3). If nuclear species (Figure 21-5). This is a thought
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 307

Relative Calcium tungstate

light
emission
Rare earth

300 400 500 600 700

Wavelength (nm)
A
Blue
sensitive film

Green
sensitive film

Relative
light
absorption

300 400 500 600 700

Wavelength (nm)
B
FIGURE 21-1  The spectrum of light emitted by radiographic intensifying screens (A) and that absorbed by a photo-
graphic emulsion (B) are familiar examples of spectra.

experiment only, so the scale is in reverse of that In any event the extremely large frequency
encountered in NMR spectroscopy. separations between nuclides and the small range
The spacing between the signals can be under- of chemical shifts for any given nuclide make it
stood in the following manner. The gyromagnetic impractical to observe more than one nuclide
ratio for hydrogen is the largest of any nuclide at a time. Therefore hydrogen signals can be
normally observed. The gyromagnetic ratio for observed completely separately from those
13
C is about four times smaller, so the carbon signals resulting from any other nuclides.
resonance frequency is reduced four times. Other It may appear from the hypothetical spectrum
nuclides scale accordingly. in Figure 21-5 that 19F and 1H are close enough
308 PART VI  Applications

0
1
2
3
Absorbance

4
5
6
7
8
9
0
340 320 300 280 260 240 220 200
Wavelength (nm)
100
Transmittance (%)

80

60

40

20

0
00

00

00

00

00

00

00

00

00

00

0
80

60
40

35

30

25

20

18

16

14

12

10

Frequency (CM−1)

CH3
OH
Signal intensity

CH2

Frequency
FIGURE 21-2  Three examples of a spectrum are ultraviolet absorption, infrared transmission, and nuclear magnetic reso-
nance emission.
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 309

RF
signal Intensity
intensity
FT

Time (s) Frequency (s−1)

FIGURE 21-3  The nuclear magnetic resonance spectrum is the Fourier transform (FT) of the free induction decay.

FIGURE 21-4  Fourier transformation is somewhat analogous to playing poker.

1
6020 H
6000
8 23
Na
Signal
intensity 6

4
14
2 N 2
H 13C 19
17
O 31
P F

0
10 20 30 40 50
MHz
FIGURE 21-5  The hypothetical nuclear magnetic resonance spectrum of the human body at 1.0 T.
310 PART VI  Applications

H
-C-H
-C
HH
H
H-O-C-C-H
HH H
-C
H

H-O

5 4 3 2 1
ppm
FIGURE 21-6  The hydrogen nuclear magnetic resonance spectrum of ethanol consists of three peaks corresponding to
the three types of hydrogen atoms according to how each is bound in the molecule.

in resonance frequency to cause accidental over- Because these electrons are moving charged par-
lap. The separation is more than 4% (megahertz ticles, they generate their own magnetic fields
apart), whereas the chemical shift ranges of each that will add to or subtract from the applied
nuclide are measured in parts per million (hertz external field. These additions and subtractions
apart). There is no overlap. are noticed by the nuclei, and the exact reso-
nance frequencies are consequently altered
slightly.
CHEMICAL SHIFT However, electrons are very small, and each
Within a single nuclear species (for example, 1H), one carries only a small charge. Their effects on
more than one peak may be present in an NMR the nucleus are small compared with the size of
spectrum. Pragmatically, it became obvious early the external magnetic field. In general, the effect
in the development of NMR that the number of of these seemingly small electronic effects is suf-
these peaks observed at low resolution was a ficient to generate a unique NMR spectrum for
measure of the number of chemically distinct every molecule.
hydrogen atoms in the molecule. For instance, A classic example of chemical shift of a single
ethanol has the chemical formula CH3CH\OH nuclear species is the hydrogen in fat and water
and exhibits three groups of signals in the NMR (Table 21-1). The hydrogen nuclei in fat are
spectrum (Figure 21-6). effectively shielded by an abundance of electrons
and therefore resonate at a 150 Hz lower fre-
The pattern of differing Larmor precession fre- quency than hydrogen in water at 1.0 T.
quencies is called chemical shift. In the molecule shown in Figure 21-6, there
are three types of hydrogen atoms, two types of
The reason for the difference in resonant fre- carbon atoms, and one type of oxygen atom.
quency for a single nuclear species is intimately This classification is based on the manner in
related to the chemical structure of the molecule which the atoms are joined in the detailed chemi-
in which that particular nucleus is bound. Each cal structure. There will be three signals in the
nucleus is surrounded by a cloud of electrons. proton (hydrogen) NMR spectrum, two in the
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 311

TA B L E 2 1 - 1 Fat and Water

0.5 T
Chemical Shift

Field Strength (T) Chemical Shift (Hz)

0.2 30
0.5 75
1.0 150
1.5 225
3.0 450 500 250 0 (Hz)
4.0 600
1.0 T

carbon NMR spectrum, and one in the oxygen

NMR spectrum.

MAGNETIC FIELD DEPENDENCE

The magnitude of the chemical shift measured in
units of frequency varies with the strength of 500 250 0 (Hz)
the external magnetic field used to obtain the
1.5 T
NMR spectrum. As the strength of the magnet
is increased, the frequency difference, and there-
fore the separation between peaks when mea-
sured in units of hertz, increases in a linear
fashion (Figure 21-7).

THE PPM SCALE

A system was adopted to simplify the reporting
of an NMR spectrum. It records the chemical 500 250 0 (Hz)
shift in a way that is independent of the strength FIGURE 21-7  The ethanol spectra of hydrogen at three
of the external magnetic field. The system used magnetic field strengths show that chemical shifts result-
to express the chemical shift is the PPM (parts ing in peak separation become more obvious at higher
per million) scale in which PPM is the difference field strength. The frequency scale is relative to a
standard.
in frequency divided by the resonant frequency
of one of the peaks, all multiplied by 1 million.

PPM (Parts per Million) if one peak resonated at 100,000,000 Hz and

another at 100,000,500 Hz, they differ by 500
PPM = (f - f0 / f0 ) ¥ 106 parts in 100 million, or 5 ppm. For 1H, most
where f is the resonant frequency under investi- resonance peaks occur over a range of 10 ppm,
gation and f0 is the resonant frequency of a whereas for 13C, resonance signals range over
standard. more than 200 ppm.

Two other features facilitate the use of the

The difference in resonant frequency between fat
PPM scale. A reference must be established, and
and water protons is 3.5 ppm.
the direction of the scale is reversed. For example,
312 PART VI  Applications

In addition, an arbitrary reference standard (3)

has been adopted for most nuclides, including H
1
H, and it is tetramethylsilane (TMS). Its unique -C-H
resonance position in each of these NMR ranges H
is given the value 0. The units of parts per million
are dimensionless but are still accorded a fre- H H (2)
quency scale. H-O-C-C-H H
Relative to the reference, the value for the H H -C-
chemical shifts can be positive or negative, H
depending on whether the resonance is at a Cumulative
higher or lower frequency than the standard intensity
(1)
TMS. As a matter of experience, as well as
H-O
choosing of a good reference, most values are
positive.
One further note is appropriate. The analyti-
cal NMR spectrometers that were in use when
these conventions were being written were oper-
ated at constant frequency, and the strength of 5 4 3 2 1
the magnetic field was altered to create a spec- ppm
trum. Thus the plot to the left of the reference FIGURE 21-8  The hydrogen nuclear magnetic resonance
spectrum of ethanol contains three peaks with sizes that
(TMS) was called upfield and the plot to the right
are proportional to the number of nuclei with equal
was called downfield. The opposite situation Larmor frequency.
occurred when the frequency swept machines
were put in wide use. The frequency is simply
plotted with increasing values to the left to
prevent two different displays. units of intensity; and the methyl (CH3) group,
three units of intensity (Figure 21-8).
Resonant peaks at higher frequency are down- The measurement protocols are more difficult
field; those at lower frequency are upfield. in the study of nuclides other than 1H, but
they can be made sufficiently precise to identify
chemical composition and structure. Because
of this, NMR spectroscopy has become one of
SIGNAL INTENSITY the prevalent methods of both quantitative and
NMR spectra are also described by the relative qualitative analysis in chemistry and biomedical
amounts of each signal type in the spectrum. This applications.
is the same as the relative number of atoms in
chemically shifted unique groups. The amount of
J-COUPLING
signal is given not by the peak height but by the
area under the curve that represents the peak. The concept and observation of J-coupling is
This relationship is readily made linear in the another critical aid to NMR. At first J-coupling
case of 1H NMR. The system is a miniature was viewed as a nuisance. However, it soon
democracy with the creed “one hydrogen atom, became a well-understood indicator of detailed
one vote.” This situation arises because the exci- molecular structure. The J-coupling appears
tation and detection of the NMR signal does not in the spectrum as additional lines appear
normally select for the chemical shift. In the case (splittings) in any particular resonance peak.
of ethanol, the hydroxyl (OH) group gives one This creates a finer, more precise spectrum
unit of intensity; the methylene (CH2) group, two (Figure 21-9).
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 313

72 66 60 54 48 42 36 30 24 18 12 6 0
Hz
FIGURE 21-9  This hydrogen nuclear magnetic resonance spectrum shows fine structure as a result of J-coupling.

The origin of J-coupling is in the interaction can be combined into imaging spectroscopy for
of spins within the same molecule. In general, medical diagnosis. Such techniques are evolving.
those spins close to each other exhibit stronger Spatially localized MR spectra can be measured
couplings than those farther away. The magni- with confidence if sufficient time and energy are
tude of the J-coupling, expressed in hertz, is devoted to the project.
independent of field strength. The effect on the
spectrum is proportional to the separation in
MEDICALLY IMPORTANT NUCLEI
chemical shift between the coupled nuclei. As a
result, high field NMR spectra are easier to inter- NMR spectral information has been used to
pret than those obtained at low field. enrich chemistry for the past 40 years; it is just
The NMR spectroscopist generates a large now attracting the interest of the medical com-
number of spectral patterns expected from struc- munity. With the application of in vivo MRS, it
tural subsets. The spectroscopist then identifies is possible to obtain various spectra from patients
structures of molecules as a whole by the inter- and therefore gain information about the chem-
pretation of these couplings. The extra informa- istry of life. Progress toward clinical utility is
tion in the J-coupling is a substantial asset in slow, but the rate of progress is increasing.
structure determination.
The figures used here to illustrate the principle
Hydrogen
of NMR spectroscopy are chosen for ease
of understanding, not for completeness of pre- The in vivo MR hydrogen spectrum is dominated
sentation. Organic chemistry and biochemistry by the signal from water. The second most
included millions of molecules that have been common contributor to the MR signal is the
studied by NMR. In most cases, the NMR spectra triglycerides found in adipose tissue (fat).
are considerably more complicated than the ones Depending on the location of the volume of
given here. However, a trained spectroscopist can interest, fat may be the only tissue present,
interpret and explain these spectra in terms of though water is usually evident. The strong (at
molecular structure and configuration. least in the NMR world) signal from the hydro-
Some words of caution: NMR spectroscopy is gen nucleus makes this nuclide the one of choice
a daunting, complex, but rewarding intellectual for MRI.
proposition, and the same can be said for MRI. MRI is relatively easy when the dominant
The expectation is that the best features of both signal is from the hydrogen in water. Image
314 PART VI  Applications

FIGURE 21-10  The observed curvilinear rim of decreased signal intensity adjacent to the renal cortex is an artifact caused
by the chemical shift between hydrogen in fat and hydrogen in water. (Courtesy George Oliver, St. Louis, MO.)

quality is compromised when fat is present. observation of at least 15 different metabolites

However, the presence of these two signals only in the brain. The MRS conditions for convenient
causes problems at high magnetic field strength. measurement select signals from N-acetyl aspar-
At low magnetic field strength the difference tate, creatine, phosphocreatine, choline, and
between the fat signal and water signal is small lactate.
enough that it is overwhelmed by the gradient The lactate is often used as an indicator of a
magnetic fields used to make the image. pathologic condition, whereas the other com-
At higher magnetic fields this difference is not pounds are found in most subjects. The database
overwhelmed by the gradient magnetic fields, for 1H MRS is accumulating rapidly. The wide-
and distinct fat and water images that are slightly spread diffusion of this database will encourage
offset from each other are produced. The result- the clinical use of MRS in the future.
ing image contains a chemical shift artifact
(Figure 21-10). The fat and water signals are
Phosphorus
10,000 times stronger than the signal from other
hydrogen-containing metabolites. The use of Another nuclide currently receiving attention in
hydrogen NMR to study these metabolites is MRS is the phosphorus isotope 31P. This nuclide
thus a strong challenge. is the only isotope of phosphorus found in nature.
Currently, the use of fat and water sup­ Phosphorus is present in all human tissue and is
pression (i.e., removing these signals from the a reporter of metabolism. It has a spin quantum
MRS region of interest) permits the noninvasive number of 1/2 and is similar in its spectral
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 315

properties to hydrogen. In other words, it is well damage in heart attacks, and monitor the effects
behaved in its MR spectral properties. of drugs and drug therapy.
One important phosphorus-containing metab-
olite is adenosine triphosphate (ATP). Others
Carbon
include adenosine diphosphate (ADP), a byprod-
uct of ATP metabolism, and adenosine mono- Almost every chemical compound in living
phosphate (AMP), a building block for the systems contains the carbon atom. Therefore it
formation of ADP and ATP. Creatine phosphate is anticipated that any method to observe carbon
(PCr), a chemical intermediate for the storage of with MR spectroscopy would be advantageous,
biochemical energy, is also evident in the 31P MR but nature has conspired to provide carbon in an
spectrum from some tissues. NMR silent form.
All of these metabolites enter into reactions in The ordinary nuclide of carbon, 12C, is
which phosphoric acid, known to physiologists nonmagnetic because it has paired nucleons
as inorganic phosphate (Pi), is either formed or and therefore does not generate an NMR spec-
consumed. One interesting application of 31P trum. The magnetic form of carbon, 13C, is
MRS is to determine the intracellular pH from present in all tissue to the extent of 1.1%. A
the chemical shift of the inorganic phosphate number of laboratories are studying this scarce
signal. Thus the MR spectrometer is a sophisti- nucleus, though the studies are experimentally
cated, expensive, but noninvasive pH meter. demanding.
An overlay of all these phosphorus signals The scarcity of 13C permits the tagging of
appears in the MR spectrum of most tissues and experimental molecules by as much as 100-fold.
provides a window into the energy state of the The tagged molecules can be followed through a
tissue. A representative 31P spectrum is shown in number of interesting and intricate metabolic
Figure 21-11. events with MRS. A representative 13C NMR
Phosphorus MR is being used to understand spectrum is shown in Figure 21-12.
and perhaps diagnose metabolic disorders, assess
Sodium
Creatine Sodium is abundant in the body, primarily as
phosphate sodium chloride and other salts. The common
isotope of sodium is 23Na; its spin quantum
number is 3/2. A spectrum of 23Na is shown in
Figure 21-13. There are indications that the MR
signal of sodium can be used to probe the intra-
molecular and intermolecular environments of
Phospho- the molecule and to report them separately.
diester
Inorganic
phosphate
γ ATP α ATP β ATP
Fluorine
Fluorine has two advantages for observation in
human tissue: its gyromagnetic ratio is nearly as
great as that of hydrogen, and the only nuclear
species is 19F, with a spin of 1/2. Atom for atom,
+10 +5 0 −5 −10 −15 it is as easy to observe as hydrogen. However, it
ppm is almost totally absent from the human body.
FIGURE 21-11  A representative 31
P nuclear magnetic Indeed, high concentrations of fluorine can be
resonance spectrum. (Courtesy Bud Wendt, Houston, TX.) toxic.
316 PART VI  Applications

200 160 120 80 40 0

ppm
FIGURE 21-12  A representative 13C nuclear magnetic resonance spectrum. (Courtesy Bud Wendt, Houston, TX.)

200 Hz

FIGURE 21-13  A representative 23Na nuclear magnetic resonance spectrum. (Courtesy Bud Wendt, Houston, TX.)

In the event a safe agent can be identified, agents to monitor blood flow, follow metabo-
which does occur, fluorine becomes a nearly lism, and understand the effectiveness of cancer
perfect tracer, or indicator, of metabolism. In a therapy regimens.
typical measurement there would be no 19F before
the measurement, the agent would be introduced,
Nitrogen
and the arrival of 19F at the volume of interest
could be monitored. Nitrogen is nearly as common in biology
In the vocabulary of imaging these are dark as carbon. All amino acids, peptides, proteins,
field measurements. There is no signal at the deoxyribonucleic acids (DNA), and ribonucleic
beginning and then an abundant signal as the acids (RNA) are rich in nitrogen. Ordinary nitro-
19
F arrives at the observation site. Figure 21-14 gen is 14N, with a spin quantum number of 1.
shows experimental 19F images focused on blood The observation of this nuclide is difficult. It has
substitute materials and pO2 imaging. an unfavorable gyromagnetic ratio and is inher-
Forms of materials that carry 19F are ently insensitive. In addition, the spectral lines
the new perfluorocarbon artificial bloods and are usually extremely broad and hard to detect.
5-deoxyfluoroglucose (5-FDG). In addition, che- A rare nuclide of nitrogen, 15N, has a spin of
motherapeutic agents for the treatment of cancer 1/2 and would seem to be suitable for study.
often carry 19F in their chemical structure. However, its gyromagnetic ratio is also unfavor-
Research is being conducted in the use of these able, and its observation is exceedingly difficult.
 CHAPTER 21  Nuclear Magnetic Resonance Spectroscopy 317

A B

C D
FIGURE 21-14  Images through the midtorso of a pig. A, Sow image. B, 19F spin echo image. C, Calculated pO2
T1-weighted (T1W) 19F image during normal breathing. D, Calculated pO2 T1W 19F image during 100% oxygen breathing.
(Courtesy Stephen Thomas, Cincinnati, OH.)
318 PART VI  Applications

If nitrogen is to be useful in medicine, it will 5. What is chemical shift as it relates to NMR

require expensive 15N-enriched substrates and spectroscopy?
long observation times. 6. What is the variation in magnetic field
homogeneity expressed in parts per million
for a 1 T ± 10 µ T magnet?
CHALLENGE QUESTIONS
7. What is the chemical shift difference in the
1. The distribution of frequencies observed in resonant frequency between fat and water
an MR signal is known as what quantity? protons?
2. List at least three other elements that could 8. What is the gyromagnetic ratio for
possibly be used to make a magnetic reso- hydrogen?
nance (MR) image. 9. What is the frequency difference for proton
3. What happens to the appearance of an NMR spins in water versus those in fat at 1.5 T?
spectrum at high magnetic field strength? 10. Along which axis does the chemical shift
4. What is J-coupling? artifact appear?
 CHAPTER

22 
Partially Parallel Magnetic
Resonance Imaging

OBJECTIVES
At the completion of this chapter, the student • Describe the acceleration factor, R, and what
should be able to do the following: it represents.
• Identify the equipment requirements for • Describe the geometry factor, g, and how it is
partially parallel imaging. related to SNR.
• Explain the difference between image based • Identify common clinical applications for PPI
reconstruction and k-space reconstruction in approaches to imaging.
partially parallel imaging.

OUTLINE
General Description of Parallel K-Space Based Parallel Applications of Parallel
Imaging Imaging Reconstruction Imaging
Image Based Parallel Imaging SNR and the Geometry Extensions of Parallel Imaging
Reconstruction Factor

KEY TERMS
G-factor R-factor Specific absorption rate
Parallel imaging Sensitivity encoding

In this chapter, the basic approaches to parallel images from the individual coils are stitched
imaging—and there are several—will be intro- together to form an overall image of a large body
duced. Parallel imaging is a relatively new type part that has both good SNR and relatively good
of MR imaging that exploits features of phased image uniformity.
array RF coils to significantly increase the speed Parallel imaging is a method for encoding the
of MR image acquisition. data which is generally compatible with many
Recall that phase array coils are groups of types of pulse sequences. It is particularly useful
receiver coils that each sense the nuclear magne- in high-field MRI systems (3 tesla and above)
tization from a small portion of the patient. The because it allows the use of fewer RF pulses than

319
320 PART VI  Applications

conventional imaging methods, thus reducing the interpolation algorithm is used to interpolate
potential for tissue heating. We shall explore the missing data in k-space before the Fourier
hardware requirements and some of the new transformation occurs. The algorithm known as
imaging parameters that are important for paral- GRAPPA is most commonly implemented on
lel imaging; particularly the relationship between clinical MRI systems now, although other varia-
the geometry factor and SNR in parallel imaging. tions of this method are under development.
We shall also take a look at various clinical You may ask, “What’s the difference between
applications of parallel imaging. A comprehen- parallel imaging and just regular imaging with
sive account of what parallel imaging can do phased array coils?” Well, in Figure 22-1 two
is not necessary for operating an MR imaging examples of the images obtained from an eight-
system, but in the description below is aimed a channel phased array coil are shown. On the left,
giving the reader a flavor of the advantages of there is a typical phased array type configuration
parallel imaging and some situations in which it in which each phased array coil picks up differ-
might be used. ent signals from different parts of the phantom
and then these images are added together to form
an image of the phantom that looks much more
GENERAL DESCRIPTION OF
uniform than you can get from any individual
PARALLEL IMAGING
surface coil.
The key elements of partially parallel imaging With parallel imaging, to speed up the imaging
(PPI) are, first of all, it uses information obtained process, data are acquired for a smaller field of
from arrays of RF coils sampling data in parallel. view. For instance, if you want to double the
Since each coil element is located at a different imaging speed, then use an FOV in phase-
position, this information can be used to perform encoding direction that is half of what you really
some portion of spatial encoding that was usually want. Obviously, this takes half of the imaging
done by the gradient fields, particularly the time but it also creates problems.
phase-encoding gradient field. A reduced data set has been acquired with
Imaging time is reduced because the phase- each coil and so, as we shall find out in Chapter
encoding process is the slowest part of image 27, the image will have a wraparound or aliasing
formation in MRI. So fundamentally PPI speeds artifact that appears in the phase-encoding direc-
up the MR image acquisition process and it does tion in each individual image where the top part
this without needing faster switching gradients, of the patient is wrapped into the bottom part of
avoiding the tissue stimulation issues they can be the image. Now, the whole trick of the parallel
caused by using additional RF pulses. In fact, imaging process is to unwrap these aliased images
parallel imaging can actually reduce the number and put them back together into one image. We
of RF pulses so that the RF power deposited is get increased imaging speed from this method,
not a big concern. but we will also have lower SNR since we are
Now there are two fundamental approaches acquiring real data for fewer lines in k-space.
to parallel imaging. The first is an image based
method in which the image is basically recon- Partially parallel imaging methods increase
structed from the signals detected by each coil imaging speed by reducing the number of
element in the phased array with a reduced field phase-encoding lines in k-space that need to be
of view and then merged. The most common acquired.
implementation of this is known as SENSE,
which stands for sensitivity encoding. A useful analogy is to compare parallel MRI
The second approach to producing parallel to x-ray CT where projections of data are used
images is the k-space based approach. In this for both cases. But in parallel MRI, the projec-
method the raw image data are analyzed and an tions come from each of the RF coils and their
 CHAPTER 22  Partially Parallel Magnetic Resonance Imaging 321

1
8 1 2
8 2

7 7 3
3

6 4 6 5 4
5

FIGURE 22-1  Left: Conventional phased array in which each phased array coil picks up a different image and all the
images are added. Right: Parallel imaging, in which a reduced data set is acquired with each coil and special processing
is used to create the composite.

spatial position and their sensitivities are a func- is practical only for designing the RF coils used
tion of spatial position (Figure 22-2), so this is a for parallel imaging. In the general clinical situ-
very simple analogy that has some limitations. ation, extra data collection on the patient, either
The first major problem is that in general there before or during the imaging, is used to gain
is no fixed geometry between the coils and the a priori knowledge of how the patient and coil
patient. Second, there is coupling between the work together before parallel imaging recon-
patient—we can think of the patient as being struction is performed.
composed of conductive bags of salt water—and
the coil array. Third, the setup changes from Parallel imaging requires that extra data on the
study to study and may even change sometimes RF coils’ signal sensitivity as a function of posi-
within a given study. tion in the patient be obtained before image
There is also the fact that the surface coils reconstruction.
themselves receive MR signals in an inherently
nonuniform fashion. These factors make it seem
like it’s going to be very difficult to determine
IMAGE BASED PARALLEL
spatial properties of the signals based on the
geometry and locations of the coils.
IMAGING RECONSTRUCTION
The key to parallel imaging is to acquire a SENSE is the primary image based form of PPI
little bit of extra data to get the information which is, in some form or another, offered on all
needed regarding the spatial distribution of RF the high-end imaging systems of all the major
coils’ sensitivity. One way to do this is by elec- manufacturers. In the common implementation
tromagnetic modeling of the coils’ properties of SENSE, sensitivity profiles are determined for
applied to phantom measurements. However, each element of the phased array coil. Measuring
because every patient differs and the way every the coil sensitivity profiles gives the system the
patient couples with a coil differs, this approach additional information necessary to constrain the
322 PART VI  Applications

3 RF coils

X-ray CT Parallel MRI

FIGURE 22-2  Diagram of analogy of parallel MRI to x-ray CT. In both cases, projections of data are used. But in CT the
tube-to-detector relationship remains fixed, which is not the case for parallel MRI.

reconstruction process so that it can be accurate too small, the SENSE algorithm is unable to
and efficient. handle it.
The two basic methods used for parallel In Figure 22-3 a reduced FOV image is shown
imaging differ in their approaches to obtaining with an aliasing artifact. Each of the four images
the data to solve the problem of recovering the on the left was acquired by one coil, and the
spatial information from the set of RF coils. SENSE process will take these four data sets and
The coil sensitivity profiles contain the data combine them into one. Because the acquisition
that describes the spatial dependencies of each uses reduced fields of view, this image is acquired
RF coil. in one quarter of the time it would take with one
The specific strategy of SENSE is to acquire coil or even with this same coil if parallel imaging
low-resolution reference images that produce reconstruction was not used.
coil sensitivity profiles and, using these data, That all sounds pretty simple, but there’s
speed up the imaging by a factor that theoreti- really a lot of stuff going on behind the scenes.
cally is as high as the maximum number of There are several steps that go into this whole
phased array coil elements and RF channels. This process. Several reference images are acquired—
“speeding-up” factor is sybolized by the letter R one for each phased array coil element—and
and is also known as the scan reduction factor also one is acquired using the integrated body RF
or the acceleration factor. When the imaging time coil. This later image acquisition is the standard
is reduced, however, the SNR is also reduced by image, because the big body coil is assumed to
at least the same factor as the speed increases, produce a pretty uniform RF field and signal.
usually more, and sometimes much more. Then the reference image data are divided up into
An important constraint with SENSE is that the data for each individual coil to get a raw
the field of view must be larger than the patient’s sensitivity map.
body in order to avoid artifacts. The system Now the problem with this approach is
is unwrapping the images during the SENSE that it messes up the noise, and particularly the
reconstruction process, and if there is additional background noise statistical characteristics. So
aliasing going on because the field of view is then a threshold mask must be generated which
 CHAPTER 22  Partially Parallel Magnetic Resonance Imaging 323

Sense

FIGURE 22-3  Left: Four reduced FOV images obtained from each of four coils. Right: The SENSE process takes
the four data sets and combine them into one. This image is acquired in one quarter of the time that is possible with
one coil.

basically removes the unwanted background

k-SPACE BASED PARALLEL
signal. Then, in some implementations, the FOV
IMAGING RECONSTRUCTION
is limited through the use of an extrapolation
zone. Finally a polynomial fitting procedure is The second approach to parallel imaging uses
calculated to produce the final sensitivity profile the raw data. The actual MRI signal acquired is
for each surface coil. This process is repeated for stored in a raw data matrix and a Fourier trans-
each of the coils in the phased array coil matrix. form algorithm is used to recover the image. This
Once the various sets of image data are stored, process is usually transparent to the MRI user,
then the reconstruction algorithm can be used to and often the system erases these data as soon as
determine the signal. This is a mathematically the transform is completed (or maybe when the
involved process which will be sketched out here next study starts) unless something special is
in very broad strokes. done to save it. So the idea of the second approach
The signal received by each coil is represented is to deal with the data in k-space before the
by an element of a matrix that is the sum of all Fourier transform is done.
the coil element sensitivities times the true trans- Another time-saving strategy is to use so-called
verse nuclear magnetization, MXY (x,y), at each autocalibration methods. Instead of acquiring
point in the patient. The gist of the math is that a whole reference image, only some extra lines
if we know what signal we measured at each in the center of k-space are acquired. A fitting
point in the patient and we know what the sen- procedure is used then to determine weighting
sitivities are for the coils at each point in the factors for generating the missing raw data lines
patient, then we should be able to recover MXY for each coil and then the Fourier transform
(x,y) at each point in the patient. This is done is used to produce an uncombined image for
through a process known as matrix inversion, each coil. With this method the data are partially
which is computationally very intensive, so there sampled and a fitting procedure is used to fill
is a noticible delay between finsihing the data in estimates of the undersampled data. Only
acquisition and display of the parallel processed after all this is the Fourier transform algorithm
images on most MRI systems. employed.
324 PART VI  Applications

FIGURE 22-4  For this patient, 2 body flex coils (4 channels each) were combined with the spine coil in the couch (up
to 10 channels available) for obtaining a high-quality thoracic/abdominal examination.

Autocalibration PPI scans include the acquistion been sampled. The fitting procedure can be per-
of extra lines of raw data during the scanning formed using multiple reference lines from each
procedure instead of acquiring entire reference of the RF coils.
images. So in a typical situation with R = 2, you may
have (24 references lines + 166 interleave data
The most common commercial implementa- lines) × 16 RF coil elements = 2240 lines of
tion of the k-space PPI method is GRAPPA, real data from which to recover the 166 lines of
which stands for generalized autocalibrating interpolated data that were not scanned. You can
partially parallel acquisition. Now, dense sam- see that there are actually more data available
pling occurs in the center of k-space, with each than is really needed to perform the reconstruc-
reference data line or the autocalibration scan tion. The trick in peforming GRAPPA is to deter-
(ACS) lines, from a single frequency-encoded mine how many reference lines you want to
signal (i.e., 100% sampling over 10% of k-space). acquire and which lines of acquired data you
However, some of the data lines acquired farther want to use to recover your missing data to
away from the center of k-space in the phase- perform the interpolation in the most effective
encoding direction are not sampled (i.e., 50% and efficient manner.
sampling over 90% of k-space). In GRAPPA the The aliasing artifacts that occur with SENSE
information in these reference lines (10% of imaging do not wrap into the edges of the images,
k-space) and the information in the lines that as occurs with aliasing in conventional MRI
were scanned at periphery of k-space (45% of methods. Instead they wrap into the middle of
k-space) are used to come up with good estimates the image, which is probably where the most
of the lines for which data weren’t actually important structures are located (Figure 22-5).
acquired (the remaining 45% of k-space). An advantage of GRAPPA is that since the algo-
This isn’t all the data that are available. There rithm does not assume anything about aliasing
are actually a number of RF coils in the array, in the image, the aliasing artifacts are much less
which are all acquiring data at the same time— severe than in SENSE. Thus GRAPPA is prefer-
anywhere from 4 to 128 RF coils (Figure 22-4). able if the clinical situation demands an imaging
The data from the other RF coils can also be used method in which the FOV is smaller than the
to help estimate the data for a line that hasn’t yet body part under investigation.
 CHAPTER 22  Partially Parallel Magnetic Resonance Imaging 325

Grappa Sense
FIGURE 22-5  Parallel MR images show more prominent aliasing artifacts in SENSE (right) compared to GRAPPA (left)
images.

SNR AND THE GEOMETRY FACTOR

was divided by the square root of R, resulting in
Because PPI uses a whole different method for a fudge factor called the g-factor (for geometry
processing MR images, it is important to con- factor), which gives the SNR for parallel imag-
sider the procedure carefully. In theory you can ing if all other imaging parameters are held
choose to accelerate the image acquisitions up constant.
to a speed that is limited by the number of inde-
pendent RF channels or coil elements available. G-Factor

However, in practice it doesn’t work out that g ( x , y , z ) = SNRNORMAL ÷ (SNRPPI × R )
cleanly, because the physics conspires against us.
In addition to the R-factor, which was dis- This g is also known as the coil-dependent
cussed previously, there is also a parameter in PPI noise amplification factor across the image
called the “geometry factor,” which relates to the volume. Engineers, who develop the phased-
coil and somewhat to the way the coil couples array coils for parallel imaging, measure the SNR
with the patient. Recall that theoretically PPI under different acceleration (R) factors and use
imaging time can be reduced by the R-factor, this equation to develop a map of the g-factor
which is equal to the number of phased array coil across the coil volume (Figure 22-7). This is
elements. However, it was observed that when important because g is spatially dependent so
this full imaging acceleration is attempted, SNR that, unlike conventional MR images, in PPI
rapidly decreases and image nonuniformities images, noise is going to change across the image
rapidly increase (Figure 22-6). In order to manage with spatial position.
these problems, the actual acceleration factor Not only does g change with R and position
must be limited to be well below the theoretical in the patient, but also it changes with the number
maximum, R. of coil elements (Figure 22-8). The more coil ele-
Pruessmann proposed the use of a “figure ments available in the phased array, the less
of merit” for PPI in which the normal SNR rapidly g decreases as R is increased.
326 PART VI  Applications

R1 R3 R4

R5 R6 R7

FIGURE 22-6  This series of images depicts reduction of SNR in an image of the heart as the acceleration factor is
increased.

Mean g(R)
5
Experimental
Fitted
4

Imaging volume
g 3

1
1 2 3 4 5 6
SENSE reduction a
R
FIGURE 22-7  The data graphed on the right depict how the g-factor increases as the acceleration factor, R, increases.
 CHAPTER 22  Partially Parallel Magnetic Resonance Imaging 327

R2 R3 R2 R3

R4 R5 R4 R5
3 3

R6 R7 R6 R7
A 1 B 1
FIGURE 22-8  The nonuniform noise distribution in parallel imaging is a function of phase-encoding orientation and
becomes increasingly noticeable at higher accelerations (R values). A, Phase-encoding direction is right to left for parallel-
encoded images acquired using R = 2, R = 3, R = 4, R = 5, R = 6, and R = 7. B, Phase-encoding is from posterior to
anterior for parallel-encoded images acquired using R = 2, R = 3, R = 4, R = 5, R = 6, and R = 7.

PPI also becomes more efficient at higher B0 is affected by which physical direction is used for
field strengths, with factor dependence on R the phase-encoding direction, since the direction
being reduced at high magnetic fields. Thus PPI used for phase encoding is the same direction in
improves as the number of useable elements in which PPI reconstruction is applied. This is true
the phased array coil increases. This has led to even for the same patient, using the same coil,
the development of premium high-field clinical with the same sensitivity factors.
MRI systems that offer options on many RF
channels, often as many as 32 channels. In PPI, noise is amplified in a spatially variant
RF channels can be quite expensive; the dif- manner that depends on the specific geometry
ference between a system with 16 RF channels of the RF coil array and is characterized by the
and 32 RF channels can be as much as $250,000 g-factor.
and coils that have many array elements are
expensive too. Therefore, care must be taken in
matching the number of channels on a coil to
APPLICATIONS OF
the number of channels a system provides. For
PARALLEL IMAGING
instance, a 32-channel coil can’t be used on a
16-channel system. Recently, at least one manu- A good example of an application for which
facturer has put the RF digitization circuitry parallel imaging is clearly helpful is the fast spin
directly on the coils, allowing their system to be echo pulse sequence, the standard T2-weighted
scaleable to coils with any number of channels. pulse sequence. You may remember that in this
In regions of PPI images with high g-factors pulse sequence a 90° pulse is followed by a train
(i.e., g > 1.5 or more) the noise will manifest as of 180° pulses. A series of spin echoes is acquired
large patchy regions across the image. This noise and each of these echoes goes to fill a different
328 PART VI  Applications

line of k-space so that the total scan time is to shorten TE so that phase dispersion does not
reduced by the number of echoes acquired per play out for a long period of time and susceptibil-
excitation. But each of these echoes is at a differ- ity is significantly reduced (Figure 22-11).
ent TE value, and therefore T2 decay affects each When should you use parallel imaging? It’s
differently. obviously useful to reduce total scan time, which
The echoes at the end of the echo train are
going to have much less SNR due to T2 decay
than those that are early in the echo train. This Blurring due to T2 decay during readout
situation causes image blurring in the phase- 102
With SENSE
encoding direction for FSE images. With PPI, a Without SENSE
conventional FSE echo train that has 96 echoes
can be reduced to 48 if PPI is used with R = 2.
This allows a reduction of the signal acquisition

FWHM (pixel)
time and results in both increased spatial resolu-
tion and greater overall SNR (Figure 22-9). 101
The penalty for using PPI with FSE is the extra
aliasing artifacts and the unique noise issues that
come with PPI. But if those problems are mini-
mized, FSE can be significantly improved using
PPI, especially at 3 T (Figure 22-10).
Another issue, particularly important to func-
100
tional imaging, is the susceptibility artifact. Using 0 20 40 60 80 100
EPI or GRE imaging, susceptibility artifacts show T2 (msec)
up in the regions of the head where the air-bone
FIGURE 22-9  With SENSE, the echo train length (ETL) is
interfaces at the sinuses cause large local gradi- reduced from 96 to 48. The time to acquire all echo signals
ents, affecting image quality. The strategy applied is reduced, resulting in improved spatial resolution in the
using PPI with these types of pulse sequences is phase-encoding direction and greater overall SNR.

Full Fourier Sensitivity

encoding Encoding

FSE

FIGURE 22-10  These head images show how sensitivity encoding improves resolution of signal-to-noise ratio. Compare
the brain image made using full Fourier encoding with fast spin echo (left) and the same procedure with an added SENSE
with a factor of 2 (right).
 CHAPTER 22  Partially Parallel Magnetic Resonance Imaging 329

A B

C D
FIGURE 22-11  Parallel imaging reduces the number of phase-encoding steps required to form an image. A and B,
Normal acquisition of brain (R = 0). C and D, Parallel imaging with R = 2 acceleration. Arrows indicate reduced suscep-
tibility induced signal loss in the temporal lobes due to shorter EPI image acquisition time with parallel imaging.

is almost always desirable. But parallel imaging of time signals have to dephase is reduced. Paral-
is also useful to speed up single-shot MRI lel imaging is also useful in any pulse sequence
methods, allowing acquisition of a whole image that incorporates a lot of RF pulses, particularly
in an even shorter amount of time. This allows 180° RF pulses, because it reduces the specific
a decrease in the signal acquisition time of echo absorption rate (SAR), which is the measure of
trains while recovering some SNR that would RF heating. This last point is particularly impor-
otherwise be lost by increasing R. tant for body imaging at 3 T where the SAR
PPI also reduces artifacts due to susceptibility limits are easily reached, which constrains the
and helps to deal with chemical shift and other operation of the MRI system due to safeguards
artifacts related to dephasing because the amount imposed for regulatory compliance.
330 PART VI  Applications

EXTENSIONS OF CHALLENGE QUESTIONS

PARALLEL IMAGING
1. How do phased array RF receiver coils
Parallel imaging has clearly been an enabling work?
technology for MRI at 3 T and above and con- 2. What portion of spatial encoding in the reg-
tinues to be an exciting area of development in ular imaging process does parallel imaging
MRI. Some systems now offer two-dimensional partially eliminate?
PPI, in which arrays of many coils can be used 3. Explain why SENSE is described as an image
to obtain massive acceleration of 3D-FT MRI based reconstruction process.
sequences that include both primary and second- 4. How are the projections of data used in par-
ary phase-encoding directions. allel imaging different from those used in
PPI reconstruction techniques also can be used x-ray CT?
to accelerate in time those imaging methods that 5. Why is acquisition of a reference image nec-
repeatedly sample the same region of tissue, such essary for SENSE imaging?
as cardiac functional studies and contrast agent 6. What are the autocalibration scan lines used
uptake studies. This approach, first dubbed un­ in GRAPPA?
aliasing by Fourier encoding the overlaps in the 7. How are the missing lines of data deter-
temporal dimension (UNFOLD) does not rely on mined in the GRAPPA reconstruction
RF coil characteristics and therefore does not method?
suffer from the g-factor based SNR losses. 8. As the acceleration factor, R, goes up, what
The concept of PPI also has been applied happens to the SNR?
to the RF excitation side of MRI, allowing 9. If the geometry factor is due to RF coil
the development of more finely tailored RF properties, why does it change with spatial
slice-selection pulses. This approach, sometimes position?
referred to as “RF shimming,” uses an array of 10. How can PPI improve fast spin echo imaging?
RF coils to produce a well-controlled excitation,
which can overcome many of the problems of RF
magnetic field uniformity that hinder effective
quantitation of MR imaging methods.
 CHAPTER

23 
Magnetic Resonance
Angiography

OBJECTIVES
At the completion of this chapter, the student • Draw both two-dimensional (2D) and three-
should be able to do the following: dimensional (3D) pulse sequences for MRA.
• Describe the flow-void phenomenon. • Explain plug flow, laminar flow, turbulent
• Discuss the origin of flow-related flow, and pulsatile flow.
enhancement.
• Distinguish between time of flight and phase
contrast magnetic resonance angiography
(MRA).

OUTLINE
Magnitude Effects Phase Contrast Magnetic Resonance
Turbulence Magnetic Resonance Angiography versus
Flow-Void Angiography Digital Subtraction
Flow-Related Enhancement Two-Dimensional Angiography
Phase Shift Effects Magnetic Resonance
Flow Measurement Angiography
Magnetic Resonance Three-Dimensional
Angiography Magnetic Resonance
Time of Flight Magnetic Angiography
Resonance Angiography

KEY TERMS
Flow-related enhancement Maximum intensity projection Pulsatile
Flow-void Moment Turbulence
Laminar flow Plug flow

331
332 PART VI  Applications

Motion affects magnetic resonance imaging

(MRI) both beneficially and detrimentally. It is
possible to measure motion and to determine
information such as myocardial velocity and Plug flow
blood flow. Vessels can also appear displaced or A Vessel wall
disappear entirely. Important structures can be
obscured by motion artifacts. Nevertheless, there
is widespread acceptance of magnetic resonance
angiography (MRA), in which flowing blood is
visualized directly. Laminar flow
The effects of motion in MRI can be separated B
into magnitude effects and phase shift effects.
Each of these produces artifacts and gives rise
to methods of motion measurement and MRA.
This chapter considers motion visualization and
measurement. R a b c d R
Pulsatile flow
The four fluid motions of blood are plug flow, C
laminar flow, pulsatile flow, and turbulence.
b
The motion itself can be complicated. Fluid
Flow rate

flow is often described as being plug flow, in

which all of the fluid moves at the same speed
a c
(Figure 23-1, A). Fluid flow can also be described
as fully developed laminar flow, in which the
speed varies in a parabolic fashion across
the lumen of the vessel. The speed is slowest at
the walls of the vessel and fastest in the center R d R
(Figure 23-1, B). Turbulence is the random Time
motion of blood in regions of discontinuity, such D
as stenosis or bifurcation. FIGURE 23-1  A, Plug flow occurs when the fluid
throughout the vessel moves at the same velocity. 
Blood flow in the arteries is pulsatile with B, Laminar flow is characterized by a parabolic distribution
regular acceleration and deceleration (Figure of velocities. Fluid near the vessel wall moves slowly while
23-1, C). The velocity profile changes from plug- fluid near the center moves faster. C, In pulsatile flow, the
like (when the fluid is accelerated) to almost periods of acceleration resemble plug flow, and the periods
parabolic (when the force on the fluid is rela- of constant flow resemble laminar flow. D, The blood flow
rate changes with time.
tively constant).
In some vessels, such as the aortic root and
the femoral arteries, the direction of flow can
reverse briefly during part of the cardiac cycle. is called a velocity profile (see Figure 23-1, A, B,
At specific points in the cardiac cycle, it is pos- and C). The changing flow rate is shown
sible to have forward flow in some parts of the in Figure 23-1, D, as a function of time for pul-
lumen and retrograde flow in others. Even some satile flow.
systemic veins have cyclic variations in flow There are two basic ways to display flowing
caused by the respiratory cycle. blood and produce an MRA. During radiofre-
A plot of the velocity of blood flow as a quency (RF) excitation, the inflow/outflow or
function of distance from the center of the vessel “flight” of proton spins results in time of flight
 CHAPTER 23  Magnetic Resonance Angiography 333

(TOF) effects. During signal sampling under the Flow direction

influence of the frequency-encoded read gradient
magnetic field (BR), spins moving in the direction
of BR experience a phase shift, resulting in phase
contrast (PC) effects.
Although physiologic motions can be complex, RF
excited
only simple motion is considered in the following
slice
discussion of motion effects. This is one reason
a variety of MRA methods are in regular use.
Different methods are better in different regions
Stationary Slow Fast
of the body.

Rapid blood flow results in dark blood; slow flow Image

results in bright blood. of slice

Depending on velocity, flowing blood can FIGURE 23-2  Stationary tissue is partially saturated by a
appear dark or bright. In general, for spin echo regular radiofrequency (RF) excitation. Fluid flowing into
pulse sequences, rapidly flowing arterial blood the slice is at equilibrium and produces a stronger signal.
The faster the flow, the greater the fraction of the slice
creates a flow-void and appears dark. Slowly
thickness in which partially saturated blood is replaced by
flowing venous blood usually appears bright. fresh blood during repetition time.
Furthermore, the appearance of blood flow
is very much a function of the pulse sequence,
gradient magnetic field intensity, and slew rate. within the slice. Stationary spins will have been
Additionally, flow-compensating gradient mag- excited by recent RF pulses and thus are not at
netic fields influence the visualization of flowing full equilibrium (Figure 23-2).
blood. Spin-lattice relaxation (T1) is the process by
Three factors contribute to flow-void and which spins give up energy to the molecular envi-
signal loss during rapid blood flow: high velocity, ronment to relax to equilibrium. The replace-
turbulence, and dephasing. Similarly, three ment of partially saturated spins by fresh spins
factors principally contribute to signal gain: can be viewed as an acceleration of the T1
flow-related enhancement, even-echo rephasing, process. Of course, the fact that the new spins
and pseudogating. are different from the old spins must be ignored;
however, that is the nature of MRI.
MAGNITUDE EFFECTS
Turbulence
The discussions of equilibrium saturation in pre-
vious chapters assumed that the proton spins The identity of spins is inferred from their loca-
were stationary. When motion is perpendicular tion. The lumen of a vessel actually contains a
to the imaging slice, it is possible for spins that lot of different spins during the course of acquir-
have experienced one or more RF pulses, and are ing the data for a magnetic resonance (MR)
therefore partially saturated, to be replaced by image, yet they all appear to be the same in the
fresh spins. image.
The fresh spins may be at equilibrium (i.e., At regions of discontinuity in the vasculature,
MZ = M0) because they have not yet experienced such as stenosis or bifurcation, flowing blood
an RF pulse. Consequently, the fresh spins trans- becomes turbulent (Figure 23-3). The result is
ported into the slice by motion can appear exceptional intravoxel dephasing and loss of
brighter than if they had remained stationary signal.
334 PART VI  Applications

Turbulence

Laminar Flow-through Flow at

flow stenosis bifurcation
FIGURE 23-3  Turbulent flow is disorderly flow that occurs in the region of a stenosis or downstream from a
bifurcation.

Flow-Void Flow-Related Enhancement

Consider the situation present in spin echo Contrary to flow-void with fast-moving blood is
imaging (Figure 23-4). To form a spin echo, flow-related enhancement (FRE), associated with
blood must be subjected to both the 90° RF slow blood flow. During multislice imaging,
pulse and the 180° RF pulse. Therefore station- slow-moving blood entering the first slice (the
ary blood appears much like the surrounding entry slice) has a different spin property from the
tissue. tissue slice.
When blood moves slowly, many proton spins Depending on the T1 of stationary tissue and
exit the imaging section and are not subjected the repetition time (TR) of the pulse sequence,
to the 180° RF refocusing pulse. The result is the tissue will be in a state of partial satura-
reduced signal intensity. tion. The TR is too short or the T1 is too long,
When blood flow is very fast, the blood or both, for tissue spins to relax to equilibrium.
protons saturated by the 90° RF pulse leave the With no flow, the stationary blood appears the
imaging section before the 180° RF refocusing same as the surrounding tissue (Figure 23-5).
pulse and no signal is emitted. The result is Suppose the slice thickness is 10 mm and
flow-void. blood flow is 1 cm/s, typical for veins. Blood
 CHAPTER 23  Magnetic Resonance Angiography 335

Spins
90° RF saturated

Image slice

1/2 TE

Slow Fast
Displaced spins

Spins
180° RF rephased

Signal intensity

Flow void
FIGURE 23-4  Flow-void occurs when fast-flowing blood leaves the imaging slice before receiving the 180° radiofre-
quency (RF) rephasing pulse.

flowing into the slice is unsaturated at equilib-

PHASE SHIFT EFFECTS
rium, whereas that of surrounding tissue is par-
tially saturated. Therefore the inflowing blood When an excited, stationary spin is in the pres-
will have higher magnetization available (MZ) ence of a gradient magnetic field, its precessional
and will emit a stronger signal (FRE). The frequency depends on its position along that
maximum FRE occurs when blood velocity is gradient magnetic field (Figure 23-7, A). Like-
sufficient to just replace the blood in the imaging wise, a spin that moves along the gradient has
slice during TR. a precessional frequency that changes with
its position along the gradient magnetic field
When blood velocity equals the slice thickness (Figure 23-7, B).
(mm) divided by TR (ms), FRE is maximum. The change in precessional frequency with
motion along a gradient magnetic field can
FRE is most visible on T1-weighted (T1W) be directly observed, if the motion takes place
images at short TR. The appearance of FRE in while the BR is energized. However, it is usually
slices deep into the volume of multislice images manifested as a phase shift that represents
reflects the parabolic shape of laminar flow and the total difference in precession between the
therefore may appear only in the center of the moving spins and stationary spins in the same
vessel lumen (Figure 23-6). voxel.
336 PART VI  Applications

Partially
saturated spins

Image slice

Short TR
No flow Slow flow

Unsaturated
spins

Signal intensity

Flow-related
enhancement
FIGURE 23-5  Flow-related enhancement occurs when blood flow is slow. Protons at equilibrium enter the slice, while
protons in adjacent tissue are partially saturated. The unsaturated blood emits a higher-intensity signal.

FLOW MEASUREMENT
imaging. Cardiac MRI (see Chapter 26) requires
It is possible to use a bipolar gradient pulse to multiple flow-compensating gradients.
sensitize the pulse sequence to motion so that the During fast imaging with very short TR, every
previously mentioned phase shift is linearly pro- slice is an entry slice and FRE is prominent. Fur-
portional to velocity (Figure 23-8). This is the thermore, because most vessels enter the imaging
basis of most flow-measuring methods in MRI. slice obliquely, the low flip angle and short TR
Table 23-1 relates the vendor acronyms used in of fast imaging intensify FRE in all vessels.
marketing their respective flow compensation The pulse sequence usually makes two mea-
pulse sequences. surements that have different sensitivities to
Flow compensation gradient magnetic fields motion along the motion-measuring direction.
are also called motion artifact suppression The velocity is then determined from the differ-
technique (MAST) and gradient moment nulling ence in phase between these two measurements.
(GMN). The term moment refers to zero Dual measurements are necessary because
order (stationary), first order (velocity), second there are other effects that produce phase shifts
order (acceleration), and third order (pulsatility in an MR image. These include chemical shift
motion). Gradient magnetic fields are tailored to differences and magnetic susceptibility differ-
null each order moment in flow-compensated ences. These sources of phase shift give the same
 CHAPTER 23  Magnetic Resonance Angiography 337

Image
slice

Image Laminar
blood
flow
FIGURE 23-6  Because of laminar blood flow, the flow-related enhancement takes on a varying appearance in multislice
imaging.

amount of phase shift in both motion-sensitive represents a nonlinear averaging of the velocities
measurements, whereas motion gives rise to dif- present in the voxel. Improvements of the method
ferent amounts of phase shift in the two measure- and clinical flow packages are available on most
ments. Thus the difference between the two imaging systems.
measurements yields phase shifts that only arise It is also possible to use intensity effects to
from motion. This phase shift difference is measure flow. For example, one method of flow
directly related to velocity in the sensitive measurement is to apply a thin saturation pulse
direction. in a plane that is perpendicular to the imaging
plane. This makes a dark line across the image.
Phase shifts can distinguish true from false If a number of images are acquired at varying
lumens in aortic dissections. delays from this saturation pulse, the dark line
moves in regions of moving tissue (e.g., flow in
Flow is usually measured by multiplying the blood vessels).
velocity of each pixel by the area of the voxel The displacement of the line during the known
that is perpendicular to the direction of motion delay between saturation and measurement can
sensitivity. This gives a reasonably accurate result be divided by the delay time to give a velocity
in vivo, although the phase shift of the pixel estimate. For example, this method has been
338 PART VI  Applications

Stationary spins

Gradient

Moving spins

Gradient

B
FIGURE 23-7  A, Stationary spins resonate at different frequencies depending on location along a gradient magnetic
field. B, Spins moving along that gradient have a constantly changing resonant frequency. This is similar to the sound
made by a trombone that is played while moving the slide.

Phase of Phase of
stationary spins moving spins

Motion compensating
gradient
FIGURE 23-8  A bipolar gradient waveform consists of two pulses of equal area but opposite polarity. Thus it does not
have a net effect on stationary spins. However, spins moving with a constant velocity gain less phase shift during the
first pulse than they lose during the second. This results in a net phase shift proportional to velocity.
 CHAPTER 23  Magnetic Resonance Angiography 339

arteries, and peripheral circulation, including

TA B L E 2 3 - 1 Flow Compensation
arteries and veins. MRA may be the only way to
Acronyms
evaluate the circulation of patients who are aller-
Vendor Name Acronym gic to iodinated x-ray contrast agents.
General Electric Flow compensation FLOW COMP
Philips Flow adjustable FLAG
Time of flight (TOF) and phase contrast (PC) are
gradient the two methods for MRA.
waveform
Hitachi Motion artifact MAST There are two ways to approach MRA:
suppression first, analyze either the intensity or phase shift
technique mechanism for sensitivity to motion. Second,
Siemens Gradient moment GMR check to see whether the data acquisition is two-
rephasing dimensional (2D) or three-dimensional (3D).
Toshiba Flow compensation FLOW COMP Because these views are somewhat independent,
Flow artifact FAST
the sensitivity mechanisms can be treated as fun-
suppression
technique
damental, whereas the appropriateness of the
data acquisition method (i.e., 2D or 3D) for each
mechanism can be secondary.

used to detect and measure retrograde flow in the

Time of Flight Magnetic
blood vessels serving the kidneys.
Resonance Angiography
The two broad categories of motion artifacts
are those related to intensity errors and those The expression time of flight (TOF) has become
related to phase shift errors (see Chapter 28). popular to describe the intensity based approach
The phase shift errors are often detected as to MRA. Consider the effect of TR on the wash-in
streaks emanating from arteries of the heart of fresh spins. The longer the TR, the larger the
chambers. These streaks are parallel to the phase- fraction of a vessel volume within the excited
encoded direction of the two-dimensional Fourier region (i.e., the 2D slice or the 3D volume) that
transform (2DFT) image. Intensity errors can contains fresh spins at equilibrium.
actually be beneficial, and they are exploited to TOF attempts to capture this idea of spin
produce MR angiograms. replacement during the interval between RF
stimulations of the excited region. It should
not be confused with the different use of “TOF”
MAGNETIC RESONANCE in positron emission tomography (PET) image
ANGIOGRAPHY reconstruction.
MRA is now routine because of the benign nature In TOF MRA, the thickness of the excited
of MRI; therefore MRA is useful as a screening region, the flip angle of the RF pulses, and the
procedure for suspected vascular disease. TR are optimized for the speed and direction of
MRA is used extensively to examine the extra- blood flow. This maximizes the contrast between
cranial circulation of the neck, especially the the blood and the surrounding tissues.
carotid and vertebral arteries. It is also useful in The wash-in of fresh spins has partially
examining the intracranial circulation, especially enhanced the blood; a number of RF pulses
the circle of Willis. within the interval of the past several T1s
Arteriovenous malformations and aneurysms have partially saturated the surrounding tissues.
are often examined by MRA. Although slow TOF MRA works better for flow that is perpen-
venous flow still poses technical challenges, MRA dicular to the smallest dimension of the excited
is used to examine all parts of the aorta, renal region (e.g., perpendicular to the slice in a 2D
340 PART VI  Applications

acquisition) because it is easier to ensure that the Naturally, the considerations of TOF MRA
fresh spins replenish the entire length of the are important in PC MRA, because a strong
vessel within the excited region. It works better signal from flowing blood is important. However,
for thinner excitation regions because it is easier it is possible to make PC MRA sensitive to
to replenish the entire imaged length of the vessel. relatively slow flow by changing the velocity
TOF MRA is best with fast, gradient echo sensitivity.
pulse sequences, which maximize signal from
blood while minimizing tissue signal. Because of
PC MRA has the potential to combine structural
short TR, tissue is partially saturated and there-
MRA with functional velocity measurements.
fore exhibits low signal intensity. Because of
short TR, FRE is maximized and vessels are
bright. The drawback to PC MRA is that it is sensi-
tive to motion in only one direction, that of
With 2D TOF, every slice is an entry slice and FRE the motion-sensitizing gradient. Thus three mea-
is maximized. surements are needed to get isotropic motion
sensitivity.
TOF MRA also works better for faster flow, This can be advantageous when the direction
because more fresh spins are moved into the of blood flow is important; however, PC MRA
RF excited region. A potential problem arises can be a drawback when only the structure
from tortuous vessels that meander within the of the vascular tree is important and when
RF excited slice because it is more difficult to TOF MRA provides adequate information in
completely replenish the saturated blood with less time.
fresh blood.
Another problem arises in areas of disturbed
flow, such as the carotid bulb, some aneurysms,
Two-Dimensional Magnetic
or downstream of a constriction or bifurcation.
Resonance Angiography
In such cases, there may be a mixture of velocities
and therefore a range of phase shifts within the The early approaches to MRA were 2D, which
voxel that reduce the brightness of blood in the remains a clinically significant approach for
region of disturbed flow by destructive interfer- screening and quick MRA examinations. The
ence. In essence, this is a reduced effective T2. 2D approach images a moderately thick slice so
Also, the higher orders of motion, such as accel- that there is high contrast between the vessels
eration present in these structures, lead to imper- and surrounding tissues. This results in a projec-
fect motion compensation. tion image of the vessel against the background
of the projected stationary tissues. This projec-
tion image is called maximum intensity projec-
Phase Contrast Magnetic tion (MIP).
Resonance Angiography The 2D approach works better with PC
The other major type of MRA uses the phase methods than with TOF methods because it is
shift mechanism of flow sensitivity. PC MRA difficult to suppress the intensity of the station-
makes two measurements with different velocity ary tissues in a thick slice at the same time that
sensitivities. The difference between the two the vessels with flow in the plane of the slice are
measurements yields an image in which the phase made as bright as possible (Figure 23-9).
shifts are proportional to velocity. For MRA pur- PC can be acquired in 2D or 3D, just like TOF
poses, pixels in the image that have significantly MRA. As with TOF MRA, PC MRA is faster in
nonzero velocities belong within blood vessels 2D and has better spatial resolution and signal-
and are shown as blood vessels. to-noise ratio (SNR) in 3D.
 CHAPTER 23  Magnetic Resonance Angiography 341

A B
FIGURE 23-9  Two-dimensional projection magnetic resonance angiography showing (A) carotid and (B) contrast-
enhanced image of the same patient. (Courtesy Errol Candy, Dallas, TX.)

projection (MIP) method (Figure 23-11). A ray

Three-Dimensional Magnetic
line perpendicular to each pixel of the MIP image
Resonance Angiography
is projected through the 3D data set, and the
The highest-quality MR angiograms are pro- value of the brightest pixel along the ray line is
duced by 3D MRA. A 3D MRA adds spatial assigned to that pixel of the MIP image.
resolution in the third dimension, thereby making MIPs can be constructed from arbitrary
it possible to detect smaller vessels that would points of view. It is possible to edit the 3D
not stand out against the background in a 2D data set before the MIP so that only a subvolume
MRA. It also produces a 3D data set from which is used for the MIP procedure. Thus it is possible
angiographic views of the vessels can be con- to engage MIP for each carotid artery individu-
structed from arbitrary directions. ally from a 3D data set of the neck.
The MIP algorithm produces MRAs that look
Three-dimensional MRA is more versatile than natural. However, it is possible for small, lower-
2D MRA. intensity details to be overwhelmed by larger,
brighter structures. Misleading depth informa-
The price for this versatility is a longer acqui- tion can also be produced, because the human
sition time. As a result, some facilities make 2D visual system often expects bright objects to be
MRAs for localization and a quick examination closer than dim objects.
in case the patient becomes uncooperative, and Among the solutions to this problem are
then they proceed to 3D examinations that can depth-weighted MIP. Depth-weighted MIP atten-
yield more information (Figure 23-10). uates the intensity of the brightest pixel along the
The angiographic view from a 3D data set ray line according to the distance between the
requires a projection of the data. The most pixel and the observer, rendering methods from
common algorithm is the maximum intensity computer graphics and stereoscopic presentation
342 PART VI  Applications

A B
FIGURE 23-10  Three-dimensional maximum intensity projection, time of flight magnetic resonance angiography
(A) midcoronal and (B) midsagittal views. (Courtesy Errol Candy, Dallas, TX.)

Projected image

2D slices
sity ber
Inten num
Slice

Patient
FIGURE 23-11  Maximum intensity projection (MIP) creates a projected image from a three-dimensional (3D) data set.
Each pixel of the MIP image has the highest intensity of pixels along a ray projected through the 3D data set.
 CHAPTER 23  Magnetic Resonance Angiography 343

A B C
FIGURE 23-12  It is possible to create separate maximum intensity projection images for each eye so that the data set
may be viewed stereoscopically. Images A and B can be viewed “wall-eyed” or with a stereoscopic viewer. Images B and
C can be viewed “cross-eyed.” (Courtesy Michael Mawad, Houston, TX.)

of the projections, which provide an improved

CHALLENGE QUESTIONS
depth perception and the appropriate relation-
ships among structures in three dimensions 1. What is the difference between laminar flow
(Figure 23-12). and plug flow?
Even with these improvements, MIP images 2. Which is more accurate in depicting blood
contain many artifacts. MIP images are good for flow in abnormal vessels: MRA or DSA?
general visualization of a lesion. However, diag- 3. What are the two principal techniques used
nosis and stereotactic surgery planning are made for MRA?
from the source images, not the MIPs. 4. What change in the MRI pulse sequence is
necessary to produce a three-dimensional
Fourier transform (3DFT) MRA image?
Magnetic Resonance Angiography 5. What is the principal problem with attempt-
versus Digital Subtraction ing to image turbulent blood flow?
Angiography 6. Which technique, 2DFT or 3DFT, produces
MRA differs significantly from x-ray contrast better spatial resolution and/or contrast res-
angiography (i.e., digital subtraction angiogra- olution in MRA?
phy [DSA]). MRA depicts flowing blood; DSA 7. If blood flow were 1 cm/s and the repetition
shows the spaces into which iodinated contrast time were 1000 ms, what slice thickness
agents eventually make their way. would produce the maximum flow-related
Thus MRA may not accurately depict the enhancements?
true lumen of a vessel because the flow is quite 8. What imaging parameters must be optimized
slow near the vessel wall. It is subject to signal for time of flight MRA?
dropout in regions of disturbed flow. MRA 9. What MRI pulse sequence maximizes flow-
may not accurately demonstrate aneurysms or related enhancement?
a thrombus. Experienced clinicians prudently 10. Diagram the flow rate of blood in the aorta
advise that raw MRA data be examined along as a function of time during the cardiac
with MIP images. cycle.
344 PART VI  Applications

BUSHONG MRI INC.

"When I told you to remove your earrings I meant

all of your rings...including your tongue ring."
 CHAPTER

24 
Perfusion Imaging

OBJECTIVES
At the completion of this chapter, the student • Identify the magnetic properties of
should be able to do the following: oxyhemoglobin and deoxyhemoglobin and
• Define exogenous and endogenous magnetic how these properties influence MRI.
resonance imaging (MRI) contrast agents. • Describe the result of BOLD imaging.
• Distinguish between perfusion imaging and
diffusion imaging.

OUTLINE
Exogenous Magnetic Endogenous Magnetic
Resonance Imaging Resonance Imaging

KEY TERMS
Deoxyhemoglobin Functional magnetic resonance Oxyhemoglobin
Diffusion imaging Perfusion

The previous chapter described the imaging Blood flows through the vascular tree into
of large vessels with magnetic resonance (MR) smaller and smaller vessels. Capillaries are the
techniques, magnetic resonance angiography smallest vessels, and they are responsible for
(MRA). This chapter describes techniques for delivering oxygen and nutrients to tissues.
functional magnetic resonance imaging, imag- Imaging blood flow in capillaries is perfusion
ing blood microcirculation, perfusion. The fol- imaging.
lowing chapter describes magnetic resonance Blood delivered to tissue, arterial blood, is
imaging (MRI) at an even lower anatomical level, oxygenated. Venous blood is removed from
diffusion. tissue through a similar capillary network as
deoxygenated blood (Figure 24-1).
Both perfusion imaging and diffusion imaging
are the bases for functional magnetic resonance Perfusion is the flow of blood through the capil-
imaging (fMRI). lary network.

345
346 PART VI  Applications

Venous blood

Tissue or organ

Molecular
perfusion
Capillary
diffusion

Arterial blood
FIGURE 24-1  Blood entering tissue is oxygenated through oxyhemoglobin. Venous blood is deoxygenated through
deoxyhemoglobin.

The x-ray tube and a hot water radiator can

BOX 24-1 Two Approaches to
serve to distinguish perfusion from diffusion. In
Perfusion Imaging
an x-ray tube, the anode heat is dissipated by
conduction to the rotor, convection from the EXOGENOUS: The contrast agent originates outside of the
housing, and radiation from the anode (Figure organ or tissue.
24-2). In a room-warming radiator, hot water ENDOGENOUS: The contrast agent is produced by manip-
is conducted through the coils of the radiator ulating the hydrogen spins within the organ or tissue.
where heat is transferred to air molecules, which
are convected through the room. In each of
15
these examples, conduction is analogous to per- O, and x-ray computed tomography (CT)
fusion; convection is analogous to diffusion. with an iodinated contrast agent provide the
There are two general approaches to perfusion basic approach to what has become exception-
imaging: the use of exogenous contrast agents ally effective MR perfusion imaging.
and the use of endogenous contrast agents The use of any exogenous contrast material
(Box 24-1). requires a bolus injection. The challenge is to
acquire images during the first pass of the bolus
of contrast material (Figure 24-3).
EXOGENOUS MAGNETIC MRI has a decided advantage over SPECT,
RESONANCE IMAGING PET, and CT in that no ionizing radiation
The use of exogenous contrast agents for study- is involved. Gadolinium chelates (Gd-DTPA) are
ing brain perfusion has been practiced for some most often used for MR perfusion imaging.
time with considerable success. Single photon MR image appearance is very much affected
emission computed tomography (SPECT) with by blood flow and bolus injection rate but, more
133
Xe, positron emission tomography (PET) with important, by the radiofrequency (RF) pulse
 CHAPTER 24  Perfusion Imaging 347

Cooling fins

Convection Conduction Radiation

FIGURE 24-2  Heat from an x-ray tube anode is dissipated by radiation, conduction, and convection. Conduction and
convection are analogous to perfusion and diffusion, respectively.

sequence. T1-weighted (T1W) pulse sequences for gray matter and white matter intensity shown
show perfused tissue with increased signal inten- in Figure 24-5.
sity. T2*-weighted (T2*W) pulse sequences result The most common perfusion protocol used
in reduced signal intensity (Figure 24-4). clinically today is T2* FID-EPI (free induction
The reversal in signal intensity is due to the decay–echo planar imaging). It is used for tumor
paramagnetism of gadolinium. In the presence characterization, tumor recurrence, and stroke.
of an external magnetic field, B0, paramagnetic
materials increase the local magnetic field because
ENDOGENOUS MAGNETIC
of the magnetic susceptibility of tissue.
RESONANCE IMAGING
The magnetic susceptibility of tissue in the
presence of a paramagnetic contrast agent causes There are several disadvantages to the use of
an intravoxel gradient magnetic field, which exogenous contrast agents. They add time and
effectively reduces T2*. The increase in local expense to any examination. If necessary, a con-
magnetization due to the paramagnetic contrast siderable delay may be required before repeating
agent increases T1 relaxation. the examination. Most important, the procedure
The reduced signal from T2*W images is invasive.
and increased signal from T1W images can be The endogenous contrast agent that has stirred
sculpted with cleverly tailored RF and gradient the most interest is blood. Arterial capillary
magnetic field pulse sequences. These effects blood carries oxygen fixed to the hemoglobin
combine to produce the time dependent signal molecule. Venous capillary blood leaves tissue,
348 PART VI  Applications

Contrast
agent
concentration

Image here

Time

FIGURE 24-3  Imaging time is critical when a bolus of contrast material is injected.

having deposited its oxygen and picked up signal intensity. Deoxyhemoglobin is paramag-
carbon dioxide. netic because it has four unpaired electrons
and therefore behaves much like the gadolinium
Arterial blood is oxyhemoglobin; venous blood is chelate contrast medium.
deoxyhemoglobin.

Oxyhemoglobin is diamagnetic and therefore Deoxyhemoglobin shortens T2*.

does not influence blood magnetization or MR
 CHAPTER 24  Perfusion Imaging 349

Normal T1W high contrast T2*W reduced contrast

A B C
FIGURE 24-4  A, Before bolus injection, there is little vascular contrast. B, After bolus injection T1-weighted (T1W)
vasculature shows high signal while tissue parenchyma is unaffected. C, T2*-weighted (T2*W) vasculature contrast is
reduced because of susceptibility effects on tissue by the contrast medium.

T2*W
signal Time (s)
intensity
10 20 30 40 50 60
0

−0.1

White matter
−0.2

Log (S/So) −0.3

−0.4 Gray matter

−0.5

−0.6 Tumor
FIGURE 24-5  Intravoxel incoherent motion results in this time-related signal intensity pattern for brain tissue.

Neuroscientists have long understood that that increases, even to the point of oversupply,
cerebral blood flow (CBF) increases in response blood flow to the active region of the brain. This
to brain activity. Regardless of whether the brain is the basis for fMRI.
activity is motor, as in tapping a finger, or visual, The hemodynamic response is not instanta-
as in viewing a picture, or aural, as in speaking, neous; it occurs over several seconds, which
each is accompanied by a hemodynamic response requires that MRI be done over a similar time
350 PART VI  Applications

Oxyhemoglobin
Deoxyhemoglobin

A B C
FIGURE 24-6  Blood oxygen level dependent imaging requires two images. A, Normal vasculature with mix of oxyhe-
moglobin and deoxyhemoglobin. B, Brain activity consumes oxygen, causing deoxyhemoglobin to form reducing signal.
C, Replacement oxyhemoglobin results in increased signal.

period. Gradient echo imaging is acceptable, but or subtraction mode. With the use of gradient
echo planar imaging (EPI) is most successful. echo pulse sequences with repetition time (TR),
The increased cerebral blood flow is im- approximately several 100 ms, baseline images
aged with fast pulse sequences tailored to deoxy- are acquired, and then the study is repeated
hemoglobin and called blood oxygen level during patient stimulation. Alternatively, with
dependent (BOLD) imaging. BOLD imaging is less than 100-ms EPI, baseline images and patient
shown schematically in Figure 24-6. With nor- stimulation images can be cycled for increased
mal brain activity, there is a mix of oxyhemoglo- temporal resolution because the cycling is less
bin and deoxyhemoglobin (see Figure 24-6, A). sensitive to patient motion (Figure-24-8).
The hemodynamic demand from brain activity Once pairs of BOLD images have been
first causes deoxyhemoglobin to form (see Figure acquired, highlighted or subtraction images are
24-6, B), which reduces signal intensity because produced from the sets of baseline and stimulus
of accelerated T2* relaxation. Next, oxygenated images (Figure 24-9). Often color is used to dra-
inflowing blood displaces the deoxyhemoglo- matically indicate areas of cerebral stimulation.
bin and signal intensity increases (see Figure The perfusion images in Figure 24-10 show
24-6, C). delayed time to peak (TTP) in the right middle
BOLD images are obtained in paired acquisi- cerebral artery area. The cerebral blood volume
tions (Figure 24-7) and displayed in highlighted (CBV) image shows normal regional CBV as flow
 CHAPTER 24  Perfusion Imaging 351

Baseline

Stimulation

Continuous multiple image

acquisition with image subtraction
FIGURE 24-7  Functional magnetic resonance images are acquired as multiple repetitions of baseline—stimulated =
highlighted image.

RFt

RFs

BSS

Bφ

BR

FIGURE 24-8  Exceptionally fast, less than 100 ms, functional magnetic resonance imaging is performed with echo planar
imaging techniques to minimize the effects of patient motion.
352 PART VI  Applications

FIGURE 24-9  Representative functional magnetic resonance imaging showing areas of cortical activity in response to
stimulation—finger tapping. (Courtesy Katey Meadors, Stillwater, OK.)

A B
FIGURE 24-10  A, Perfusion images of regional time to peak (TTP). B, Perfusion image of cerebral blood volume (CBV).
(Courtesy Todd Frederick, Dallas, TX.)
 CHAPTER 24  Perfusion Imaging 353

equilibrates over time. Such “brain maps” are 5. Which has shorter T2 relaxation time: oxy-
very effective perfusion studies for patients with hemoglobin or deoxyhemoglobin?
a stroke. Such images are usually displayed in 6. SPECT and CT can produce excellent func-
color. tional images. What is one significant advan-
tage of fMRI over these other techniques?
7. Describe the BOLD sequence for fMRI.
CHALLENGE QUESTIONS
8. Why is fast imaging required for BOLD con-
1. What is the difference between perfusion trast fMRI?
and diffusion? 9. What is the importance of imaging time for
2. What does BOLD stand for? exogenous fMRI?
3. What type of MRI pulse sequences are used 10. What is the effect of deoxyhemoglobin on
for fMRI? proton density, T1 relaxation time, and T2
4. Distinguish between exogenous and endog- relaxation time?
enous as the terms are applied to MRI con-
trast agents.
 CHAPTER

25 
Diffusion Imaging

OBJECTIVES
At the completion of this chapter, the student • State the signal attenuation equation and
should be able to do the following: identify each parameter.
• Identify brownian motion and its relationship • Identify two magnetic resonance imaging
to diffusion. (MRI) pulse sequences used for diffusion
• Discuss the parameter, b, and its place in imaging.
diffusion imaging.

OUTLINE
Tissue Diffusion Pulse Sequences

KEY TERMS
Diffusion coefficient Signal attenuation
Diffusion gradient Slew rate

Diffusion is one level closer to the cell than capil- required for molecules to diffuse from one
lary perfusion. Diffusion is the random motion medium to another is expressed by the diffusion
of molecules from a region of high concentration coefficient, D.
to one of low concentration. Consider the situation shown in Figure 25-1.
If one could isolate two different molecular
species into compartments, with time, the mole-
TISSUE DIFFUSION cules would mix. The rate at which such mixing
Diffusion is also termed brownian motion occurs is best described by D.
for Robert Brown, the English scientist who If the diffusion is restricted, as occurs across
first identified this process as thermal, or heat a cell membrane, the diffusion distance is limited
induced. At T = 0 K, absolute zero, there is no (Figure 25-2). These relationships are described
diffusion. As temperature increases, molecules mathematically by Fick’s law and Einstein’s
in one medium vibrate, move, and diffuse into equation, from which one can derive D. The dif-
an adjoining medium more readily. The time fusion coefficient has units of mm2/s.

354
 CHAPTER 25  Diffusion Imaging 355

Membrane
FIGURE 25-1  Diffusion occurs when different fluid molecular species mix. The time required for complete mixing is a
function of the diffusion coefficient, D.

Free
diffusion
Diffusion distance (µm)

Cell membrane

Restricted diffusion

Diffusion time (ms)

FIGURE 25-2  Diffusion is slower and limited when a barrier, such as a cell membrane, is present.
356 PART VI  Applications

90° 180°
RFt

SE

RFs

BD
Tag Untag
Diffusion time
Stationary spins
FIGURE 25-3  Spin echo diffusion imaging requires that a gradient magnetic field be pulsed briefly on either side of the
180° RF refocusing pulse.

RFt

90° 180°
SE

RFs

Diffusing spins
FIGURE 25-4  Spins diffusing during the time between the diffusion gradient magnetic fields experience a phase shift.

Higher values of D represent faster rates of angiography (MRA). The simplest pulse sequence
molecular diffusion. is spin echo (Figure 25-3). The three gradient
magnetic fields are produced for spatial local-
The diffusion coefficient, D, has been mea- ization. Additionally, one of the gradients is
sured for many body fluids. The reciprocal of D, energized briefly on either side of the 180° radio-
a value termed b that depends only on gradient frequency (RF) refocusing pulse. This is called a
magnetic fields, is that which is computed for diffusion gradient.
magnetic resonance (MR) images. Therefore the Now, consider the response of intravoxel spins
parameter b has units of s/mm2. due to diffusion during the period between the
The diffusion imaging scheme is not unlike two diffusion gradient pulses (Figure 25-4). Spins
that for phase contrast (PC) magnetic resonance moving randomly (diffusing) during application
 CHAPTER 25  Diffusion Imaging 357

Strong echo

Spins in
No diffusion phase

Voxel

Weak echo

S0

Spins
Random diffusion dephased Echo attenuation

BD

FIGURE 25-5  The quantity S/S0 is the diffusion signal attenuation, A.

of the gradient have their transverse magnetiza- gradient may best be described as an indicator of
tion, MXY, shifted in phase. At the time of the membrane integrity.
second diffusion gradient, this phase shift results
in accelerated dephasing of the spin echo so that
PULSE SEQUENCES
the intensity of the echo signal, S0, is reduced to
S as shown in Figure 25-5. For a diffusion image to be made, two or more
The ratio of S/S0 is called signal attenuation signal acquisitions are required. Although previ-
and given the symbol A. The value of A is a ously described for spin echo imaging, nearly
simple exponent in D and b. any pulse sequence can be used: steady state free
precession (SSFP) stimulated echoes, gradient
Signal Attenuation echoes (GREs), and echo planar imaging (EPI).
Subtraction of the two or more images results
A = S/S0 = e - bD in a value for each pixel related to b. This value
is calculated by regression analysis of each image
D is a property of the fluid, and b depends to the exponential equation for signal attenua-
on the nature of the diffusion gradients (Figure tion given previously.
25-6). If there were no diffusion, the diffu-
sion gradients would have no effect and a diffu- Each image acquired during a different diffusion
sion image would not be possible. The diffusion gradient will have a different value for b.
358 PART VI  Applications

−0.1

−0.2

−0.3
S/S0 = e−bD
Log (S/S0)
−0.4

−0.5

−0.6

−0.7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
b (x 1000 s/mm2)
FIGURE 25-6  Diffusion signal attenuation as a function of b values.

A B

C D
FIGURE 25-7  A, Fluid attenuated inversion recovery (FLAIR) transverse image. B, Diffusion-weighted image (DWI) trace
image with a b value of 1000. C, Apparent diffusion coefficient (ADC) map of DWI image in B. D, Time-to-peak (TTP)
perfusion map. (Courtesy Todd Frederick, Dallas, TX.)
 CHAPTER 25  Diffusion Imaging 359

The resulting diffusion image shows fast dif- lengthened T2. The wedge shaped area on the
fusion as bright and slow diffusion as dark. TTP perfusion image represents a larger area of
However, there are many confounding obstacles delayed perfusion.
to successful diffusion imaging. Nevertheless,
while the value of b differs with diffusion gradi-
ent intensity, the value of D is a constant tissue
CHALLENGE QUESTIONS
related property.
The emergence of EPI has made rapid diffu- 1. What is a diffusion coefficient?
sion imaging possible. The improved gradient 2. How is the diffusion coefficient measured?
coil systems associated with EPI allow a more 3. Why do we use the b factor rather than the
accurate determination of D. Gradient field diffusion coefficient when evaluating diffu-
strengths of 40 mT/m with slew rates of 60 T/m/s sion imaging?
allow imaging of b factors to 500 s/mm2. More 4. How does rapid perfusion appear, compared
intense gradient magnetic fields produce higher with slow perfusion?
b values or the same b value at reduced time-to- 5. How would the diffusion gradient be identi-
echo (TE). fied on an MRI pulse sequence diagram?
This cannot be done on slower, weaker MRI 6. What is the ratio of S/S0 called in diffusion
systems. One significant disadvantage of older imaging, and what is its meaning?
systems is that the slow data acquisition allows 7. State the equation for signal attenuation and
for inadvertent motion. identify each parameter.
Combining perfusion/diffusion images is illus- 8. When the b factor is increased after increas-
trated in Figure 25-7, where there is a perfusion- ing the amplitude of the diffusion gradient,
diffusion mismatch. The small spots that are what happens to signal attenuation?
bright on the FLAIR and DWI images and dark 9. What is brownian motion?
on the ADC image represent dead tissue due to 10. What are the two principal MRI pulse
cytotoxic edema causing restricted diffusion and sequences used for diffusion imaging?
 CHAPTER

26 
Cardiac Magnetic
Resonance Imaging

OBJECTIVES
At the completion of this chapter, the student • Discuss the property of segmented k-space.
should be able to do the following: • Identify the four principal imaging planes for
• Identify the magnetic resonance imaging cardiac anatomy.
(MRI) system specifications necessary for • Discuss the process of imaging myocardial
cardiac MRI. perfusion.
• Describe the principal pulse sequences used
for cardiac MRI.

OUTLINE
Imaging System Evaluation of the Heart
Requirements Cardiac Anatomy
Imaging Techniques Myocardial Perfusion
Pulse Sequence Selection Magnetic Resonance
Filling k-Space Angiography

KEY TERMS
Breathhold imaging Myocardial tagging Segmentation
K-space segmentation Navigator echo Tissue tagging

Heart disease is one of the leading causes of an accepted method for a number of cardiac
death in America for both men and women. Any diseases. The types of tissue involved and the
improvement in early diagnosis would surely variation of motion in the heart make it tough
improve survival and quality of life. on MRI. Nevertheless, cardiac MRI is helpful in
Advances in magnetic resonance imaging evaluating both congenital and acquired heart
(MRI) system hardware and pulse sequence disease.
development are resulting in significant ad- There are three types of tissue in the heart:
vances in cardiac MRI, and it is now becoming muscle of the myocardium, fat, and rapidly

360
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 361

can be detected, and wall motion abnormalities

TA B L E 2 6 - 1 Approximate Values of MRI
under stress occur before changes in the electro-
Parameters for Heart Tissue
at 1.5 T
cardiograph (ECG) or development of chest pain.

Tissue T1 T2 PD End expiration breathhold results in less motion

artifact.
Myocardium 760 40 0.8
Fat 290 50 0.9
Blood 700* 160* 0.9 The main obstacle to adequate cardiac MRI
is respiratory motion. With every breath, the
*Very flow dependent. heart moves up 3 cm vertically. Fortunately,
breathhold images with good signal intensity and
high resolution can be obtained with short
flowing blood. Each of these tissues has suffi- breathholds of up to 16 seconds.
ciently different values of T1, T2, and proton When attempting breathhold imaging, the
density (PD) that cardiac imaging should be easy technologist should carefully instruct the patient
(Table 26-1). and direct the patient to practice. Breathholding
However, motion gets in the way to compli- at the end of inspiration results in considerable
cate MRI. Movement of the diaphragm, the variability in heart positioning; breathholding
beating heart, and flowing blood result in very at the end of expiration is preferred. Because
complex motion. not all patients can hold their breath, even for
Flowing blood within the coronary arteries a short time, other methods have been devel-
can be imaged with the techniques of magnetic oped to reduce artifacts caused by respiratory
resonance angiography (MRA). However, the motion.
gold standard remains x-ray coronary angiogra-
phy because it has better spatial resolution and Navigator echoes are fast, one-dimensional
contrast resolution. images used to monitor the position of the
Myocardial perfusion can be imaged with the diaphragm.
contrast agent gadopentetate dimeglumine (Gd-
DTPA) or with the endogenous BOLD (blood The navigator echo is acquired immediately
oxygen level dependent) technique, although the after each triggering of the ECG. If the diaphragm
BOLD technique is largely restricted to research. has moved more than a predefined distance, the
Cardiac MRI contrast studies are also useful for signal is rejected and another signal is acquired
determining which part of the myocardium is at the same phase-encoding step.
dead, or infarcted. Although navigator echoes allow good cardiac
Radioisotope imaging with thallium (210Th) images to be obtained while the patient is breath-
and/or technetium (99mTc) under stress remains ing freely, a lot of signals are rejected. The total
the diagnostic test of choice for assessing myo- image time can take two to four times longer
cardial ischemia, but MRI studies are increas- than it otherwise would.
ing as SPECT studies become less common.
Furthermore, MRI can be used to acquire mul-
IMAGING SYSTEM REQUIREMENTS
tiple images with good spatial resolution in a
heartbeat. The rapid motion of heart tissue and the move-
Wall motion imaging with spatially selective ment of blood place exceptional demands on the
presaturation radiofrequency (RF) pulse tech- function of all components of the MRI system.
niques called myocardial tagging shows great Fast image acquisition requires precisely tuned
promise for evaluation of myocardial ischemia. electronics, robust gradient coils, and a high B0
Wall position displacement as little as 1.0 mm field intensity.
362 PART VI  Applications

Several manufacturers have developed MRI

IMAGING TECHNIQUES
systems specifically for cardiac MRI. These
systems have short bore design, high B0 field There are two general approaches to cardiac
intensity, and strong, fast gradient coils. MRI. ECG-gated, spin echo (SE), and gradient
echo (GRE) techniques were the first applications
Minimum requirements for good cardiac MR for cardiac imaging and continue to be the
imaging are 1.5 T, 15 mT/m, and 100 T/m/s. standard.
The quality of the ECG gating signal must
Although most MRI systems advertise cardiac ensure quality images. Proper placement of
MRI capacity, several hardware features are nec- the ECG electrodes and the positioning of the
essary for good cardiac MRI. The following ECG cables are necessary. No loops are allowed.
characteristics are essential if the full ability for Figure 26-1 illustrates proper ECG electrode
cardiac MRI is to be achieved. placement.
Static Magnetic Field.  Although cardiac MRI Pulsing gradient fields can significantly distort
can be done at B0 as low as 0.2 T, 1.5 T is prob- the ECG signal during imaging. It is recom-
ably the minimum that should be used. The mended that the three or four ECG lead wires be
improved signal-to-noise ratio (SNR) at 1.5 T braided to reduce these effects. Also, check to
allows for faster imaging and high temporal reso- ensure that the signal is gating on the R-wave
lution, even at high heart rates. The temporal and not the T-wave for optimal image quality.
resolution at lower field intensity is too low for
state-of-the-art physiologic evaluation. Faster imaging within a single breathhold pro-
Field of View (FOV) and RF Coils.  A surface vides improved temporal resolution.
coil complement should be available. Dedicated
phased array cardiac coils, typically with four to ECG gating allows the periodic heart motion
six elements, allow for further improvements in to be stopped, like a strobe light at a disco dance.
SNR and contrast resolution. Also, in selected In conventional multislice imaging, TR is long
instances, they allow use of a small FOV (down and echo time (TE) sufficiently short so that a
to 10 cm) for detailed imaging, though there is single acquisition for each slice is obtained within
reduced coverage because only selected elements one TR. ECG-gated imaging is similar except the
of the array are used. TR approximates the cardiac cycle length.
Gradient Coil Capacity.  The gradient coils must Most patients have heart rates between 60 and
be fast and intense. Switching times of not more 120 beats/min. These rates correspond to TRs of
than 200 µs are required. When energized, the 1000 ms and 500 ms, respectively.
coils must produce a gradient magnetic field of
at least 15 mT/m but even higher is better for Question: If the cardiac frequency is 60 beats/
most applications. min, what is the time of the required cycle
and therefore the required TR?
Slew rates of at least 100 T/m/s are desirable. Answer: 60 beats / min
= 60 beats / 60 s
The fastest gradients now allow imaging = 1beat /1000 ms
with repetition time (TR) as low as 2 ms. With hence: TR = 1000 ms
single-shot echo planar imaging (EPI) techniques,
a 64 × 64 matrix image can be acquired in Because of the short TR, such images are
approximately 50 ms. This limitation is due to heavily T1 weighted (T1W). A common approach
the T2 of the myocardium being only approxi- is to make the images more T2 weighted (T2W)
mately 40 ms. Multishot EPI may be used for by skipping one or more heartbeats and increas-
better spatial resolution. ing the TR accordingly. Of course, the TE also
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 363

Surface coil on skin

Closed loops

Looped
ECG leads

Surface coil
No skin contact

Straight
ECG leads
No closed loop

FIGURE 26-1  Proper electrocardiograph (ECG) electrode placement for cardiac-gated imaging. Closed conductive loops,
even of anatomy, should be avoided.

needs to be lengthened; otherwise, we will have 4 ms), and a 20° to 30° flip angle. Images are
a long TR, short TE sequence that results in a acquired through the cardiac cycle, but only one
proton density–weighted (PDW) image. signal for each image is acquired per cycle.
The R wave is chosen to gate the image acqui-
sition because it has the highest voltage. There-
Pulse Sequence Selection fore if a 256 × 256 matrix is desired for each
The type of cardiac pulse sequence chosen for image, 256 heartbeats will be required (Figure
a given patient is determined by the objective 26-2). If the heart rate is 60 beats/min, then just
of the study. Cardiac anatomy is best evaluated over 4 minutes is required.
with SE MRI. Cardiac function is best evaluated The cardiac cycle can be sectioned into mul-
with cineradiography (cine) MRI. Myocardial tiple phases. The number of lines of k-space
perfusion requires fast GRE or hybrid EPI acquired during each phase of the R-R interval,
sequences. equivalent to each cine frame, corresponds to the
MRA is usually done with time of flight (TOF) segmentation of k-space. Segmentation effec-
techniques, though velocity-encoded phase con- tively reduces the number of cardiac cycles and
trast (PC) is sometimes used to measure blood imaging time proportionately.
flow in the coronary arteries or aorta.
Cine MRI.  A cine loop of one or more cardiac The larger the number of lines of k-space col-
cycles uses spoiled GRE or fast imaging with a lected in each segmentation, the faster k-space
short TR (less than 10 ms), a short TE (2 to can be filled.
364 PART VI  Applications

Heartbeat #1 Heartbeat #2 Heartbeat #256

Image here Image here

Signal Signal

FIGURE 26-2  An electrocardiograph of the cardiac cycle triggers magnetic resonance imaging at each R-wave. A number
of heartbeats equal to the number of k-space lines are required for each image.

There is a penalty for increasing k-space seg- However, EPI is compromised by chemical shift,
mentation to reduce imaging time. Temporal T2*, and field inhomogeneities. Nevertheless,
resolution is reduced, resulting in image motion EPI sometimes is applied with success for cine
blur. The number of cine frames for analysis is cardiac MRI.
reduced. Multislice SE.  Gated multislice SE imaging is
However, a sequence called steady state free useful for imaging cardiac anatomy. Imaging
precession (SSFP) has been used that has short times of 2 to 4 minutes will allow 10 to 15, 256
TR (2 to 4 ms) and TE (1 to 2 ms), allowing × 256 matrix images to be acquired. Such pulse
k-space to be covered quickly. This sequence sequences are a solid addition to cine MRI.
overcomes some of the previous disadvantages Individual slices can be obtained during a breath-
of cine MRI by allowing both high temporal hold with segmented k-space techniques. When
resolution and high spatial resolution (Figure combined with dual inversion-recovery pulses,
26-3). the moving blood is nulled, and crisp details of
Echo Planar Imaging.  EPI is capable of cardiac anatomy can be obtained.
extremely fast imaging; 50 ms per image or faster Spatial presaturation pulses allow the exclu-
is easily obtained. EPI also has a high SNR. sion of signal from various tissues. If many thin
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 365

FIGURE 26-3  These six segmentation-gated images were obtained in six heartbeats and can be merged to form a cine
loop. (Courtesy Geoff Clarke, San Antonio, TX.)

spatial presaturation pulses are applied to the prohibitive amount of time compared with the
myocardium right after the ECG trigger, they will faster cine methods.
persist for many phases of a cardiac cine sequence. There is a technique for “black-blood cine
This method is called tissue tagging. imaging” that uses two inversion pulses before
When the presaturation lines are applied in acquisition of the cine data. The first slice-
such a way to produce a grid of saturated tissue, selective pulse inverts all of the spins, and the
the motion of the myocardium can be followed second inversion pulse realigns the spins along
through a good portion of the cardiac cycle, and the positive BZ axis.
regions of poor function can be visualized. However, the blood in the heart chambers
moves into other slices between the two 180°
pulses and does not receive the second inversion
When compared with the myocardium, signal
pulse. After an appropriate inversion time, the
from blood using fast gradient echoes appears
signal from the blood will be nulled, whereas the
very bright with all of the faster imaging methods.
signal from the myocardium will be strong.
The drawback to this method is that the two
One of the nice things about conventional inversion pulses and the inversion delay take
SE imaging is that the blood is black, allowing additional time; often the number of cine phases
easy discrimination of the endocardial mar- is significantly reduced compared with white
gins. However, conventional SE often takes a blood cell imaging.
366 PART VI  Applications

signal averages, 256 phase-encoding steps,

Filling k-Space
3000 ms TR, and 100 ms TE?
The principal time constraint in cardiac MRI is Answer: T = 3000 ms × 2 × 256
the time required to fill k-space, which, during = 1536
, , 000 ms
SE imaging, is the product of TR, the number of = 1536 s
signal acquisitions (NSA), and the number of = 25.6 min
phase-encoding (BΦ) steps. Reduction in any one
of these parameters speeds imaging. The manner in which k-space is filled has
been exploited to reduce imaging time. Instead
of sequential filling, line by line, as with SE
Spin Echo Imaging Time imaging, half Fourier imaging (HFI), segmenta-
T = TR ¥ NSA ¥ # BF tion of k-space, and spiral filling have been
applied to fill k-space faster.
Half Fourier Imaging.  k-Space exists in four
Question: What is the time required to obtain a distinct quadrants, each of which is related to the
T2W image with the following technique: 2 other (Figure 26-4). Each line in quadrant A, for

A B
1

128 KX (frequency)
C D

254
255

256

KY (phase)
FIGURE 26-4  The four quadrants of k-space are mathematically related so that the remaining three can be computed
from the acquisition of one quadrant.
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 367

Heartbeat #1 Heartbeat #2 Heartbeat #64

Frame # Frame # Frame #
123 16 123 16 123 16

TR

Frame #1 Frame #16 Frame #1 Frame #16 Frame #1 Frame #16

FIGURE 26-5  Single-slice k-space segmentation speeds image acquisition for cine imaging.

instance, is mathematically related to the line in GRE imaging. This is still not fast enough for
each other quadrant, usually as a mirror image. routine cardiac MRI.
If one acquires only the first half of each With improvements in gradient coil design
signal, one can compute the other half, which and performance, spoiled GRE and segmented
speeds imaging a little. This form of HFI still k-space reduce imaging time to a single 20-second
requires that 256 signals be acquired. breathhold. TRs of 10 to 20 ms and TE of 2 to
A faster form of HFI requires that only 128 7 ms with flip angles not exceeding 20° are
signals be acquired, and the other 128 signals are routine. These imaging techniques result in satu-
computed to fill all 256 lines in k-space. Imaging ration of stationary tissues and flow-related
time is nearly halved because usually in addition enhancement from inflowing blood.
to the 128 signals acquired, an additional 8 to Segmentation of k-space is shown in Figure
12 central lines are acquired to maintain good 26-5 for a single slice. Each R-R interval
SNR. is divided into several periods of the cardiac
k-Space Segmentation.  For the improvement cycle. For a heart rate of 60 beats/min, 20 differ-
of temporal resolution and reduction of motion ent segments of k-space can be measured in a
artifacts, k-space can be filled in segmented 20-second breathhold. During this time, multiple
rather than sequential fashion. For k-space seg- cine frames can be acquired to produce the
mentation, multiple phase-encoded signals are cine loop.
acquired during each heartbeat. The complete
k-space is filled over 10 to 20 heartbeats, with
Cineradiography modes are useful for evaluation
an effective temporal resolution of approximately
of the myocardial function, ventricular volume,
50 ms, which allows cine acquisitions.
ejection fraction, wall thickening, and blood flow
The 4 to 7 minutes required for SE cardiac
abnormalities.
MRI is reduced (3 to 5 minutes) with ECG-gated
368 PART VI  Applications

Heartbeat #1 Heartbeat #2 Heartbeat #64

Frame # Frame # Frame #
123 16 123 16 123 16

TR

Frame #1 Frame #16 Frame #1 Frame #16 Frame #1 Frame #16

FIGURE 26-6  Spiral filling of k-space places more data in the center, which improves contrast resolution.

Segmentation of k-space can be implemented Spiral imaging has not yet received widespread
to produce multiple image planes. In this segmen- application because of all schemes it requires
tation mode, each segment corresponds to a dif- special image processing. Spiral imaging was
ferent slice, which allows an image of the entire actually conceived as a method for imaging
heart in 20 seconds. quickly, using gradients with conventional ampli-
The product of the number of lines of k-space tudes and ramping rates.
in the segmentation and the number of cardiac As such, it showed promise. However, gradi-
cycles determine the matrix size of the image. For ent amplifier technology soon allowed imaging
instance, if four lines are acquired in each cycle, systems to create fast ramping, powerful gradient
over a 20-second acquisition, the image matrix fields that were useful for other things (like dif-
will be 80 × 80. fusion weighting), in addition to fast imaging.
Spiral Filling.  An alternative to sequential Spiral imaging has not really found a foothold in
or segmental filling of k-space is spiral imaging any significant clinical application.
(Figure 26-6). Spiral k-space filling uses si­
multaneous energization of the phase- and
EVALUATION OF THE HEART
frequency-encoding gradients immediately after
RF excitation. For most examinations, 5 to 10 mm slice thick-
The two gradients begin oscillating at low ness is adequate. With thinner slices, reduced
amplitudes, 90° out of phase. The amplitude of partial volume artifact results in improved con-
these gradients increases during signal acquisi- trast resolution. However, in the heart, slice
tion. Filling begins at the center of k-space and reduction may not produce the benefits found in
spirals outward. other organs because the heart may be moving
in the slice-select direction during imaging.
With spiral imaging, the center of k-space is filled With a patient supine in a horizontal B0 field,
more than the periphery. This contributes to the axes of the heart and left ventricle are not
improved contrast resolution. parallel to the axis of the body or the B0 field. As
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 369

A B C D
FIGURE 26-7  These images show the four basic views of the heart from left to right. A, Two-chamber long axis.
B, Short axis. C, Four-chamber long axis. D, Left ventricular outflow tract. The upper row demonstrates the “black blood”
spin echo technique, and the lower row demonstrates the corresponding “bright blood” gradient echo technique (in
these images with the steady state free precession method). In the bottom row, the white arrowhead identifies the left
ventricle, and the black asterisk shows the right ventricle. (Courtesy Eric Douglas, Houston, TX.)

a result, oblique images are necessary for proper anatomical abnormalities. Ventricular volumes,
coverage of the heart. myocardial mass, and myocardial wall thickness
are three of the most important measurements
for staging congenital heart disease and conges-
Cardiac Anatomy tive heart disease and are best evaluated with
Evaluation of cardiac anatomy is best achieved cine MRI techniques.
with the segmented k-space techniques previ- Regardless of the pulse sequence used to dem-
ously described. Best results are obtained with onstrate anatomy, four anatomical planes must
a dedicated phased array cardiac surface coil be imaged (Figure 26-7).
with low bandwidth (e.g., ±32 kHz or less) for Evaluation of ejection fraction usually requires
better SNR. This reduces the TR and TE and a complete set of short axis views covering
allows longer segments of k-space to be filled the left ventricle from the base, at the mitral
each cardiac cycle. valve plane, to the apex. Segmental wall motion
Multislice or single-slice breathhold SE MRI is also best evaluated with short axis views
is best for evaluating the many types of cardiac (Figure 26-8).
370 PART VI  Applications

B
FIGURE 26-8  These short axis views demonstrate (A) black-blood technique acquired in a single breathhold with single-
shot fast spin echo and (B) bright-blood technique with a steady state gradient echo acquisition in a single breathhold.
Note the regurgitation (black jet) at the mitral valve. (Courtesy Bill Faulkner, Chattanooga, TN.)

There are three long axis cardiac views. the four-chamber view, which is useful for
imaging the mitral and tricuspid valves and is an
added measure of evaluation of segmental func-
The first long axis view is of the left ventricu- tion. The four-chamber view should pass through
lar outflow tract (LVOT), which includes the the ventricular apex and the mitral and tricuspid
aortic and mitral valves and the apex of the left valves.
ventricle. The second long axis view is the Taken collectively, these four views (short
two-chamber view, which provides additional axis plus three long axes) provide images that
images of left ventricular function and mitral allow assessment of the most important cardiac
valve function. The third long axis view is structures and function. Cardiac MRI relies on
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 371

specific measurements from the image data more However, because of arterial pressure and diffu-
than most other clinical applications. sion, some contrast agent is eventually taken up
The volume of the left ventricle can be deter- by dead tissue.
mined from measurements of the cross-sectional The contrast material also washes out much
areas of the chamber of the left ventricle, obtained more slowly from the infarcted region than
from a series of short axis views. When these from the healthy tissue. Thus if a T1W image is
measurements are derived from a set of cine obtained during a period ranging from 10 to 30
image data, the left ventricular ejection fraction minutes after injection of the contrast agent, the
can be calculated as the difference between the infarcted tissue will appear bright while the
ventricular volume at end-diastole and end- normal myocardium will remain dark.
systole divided by the end-diastolic ventricular Magnetic resonance contrast agent imaging is
volume. useful in evaluating abnormal wall motion. Wall
The wall thickness can be measured in motion analysis is effective for detection of isch-
each short axis image, circumferentially around emic and infarcted myocardium. Dobutamine-
the ventricular chamber. The percent wall thick- induced cardiac stress MRI has been shown to
ening is then taken as the difference between be better than ultrasound or nuclear medicine
the end-diastolic and end-systolic wall thickness imaging for detecting myocardial wall thickening
divided by the end-diastolic wall thickness times and dysfunction.
100%. Several groups have shown that with very
high-resolution MRI, the wall of arteries can be
Myocardial mass can also be determined by mul- well visualized. In the case of arteriosclerosis,
tiplying the total myocardium volume by the individual plaques can be identified, and the
density of the myocardium tissue, 1.05 g/cc. characteristics of the plaques, whether they are
mostly fibrous or fatty, can be determined. This
High field strength MRI systems now include has led to much excitement about the possibility
vendor-developed cardiac computer programs of using MRI to evaluate the “vulnerable plaque,”
to semiautomatically analyze for ejection frac- that is, a plaque that will soon rupture and
tion, chamber volume, and segmental myocardial thereby initiate events that will lead to a block-
function based on wall movement. age of the artery.
The most convincing images of vulnerable
plaques thus far have been in the carotid artery,
Myocardial Perfusion which is relatively motionless and superficial.
EPI with a bolus administration of the paramag- However, investigators have demonstrated the
netic contrast agent Gd-DTPA is fast becoming ability to measure the wall thickness of the
the technique of choice for evaluating the perfu- coronary arteries, too. Thus it is likely only a
sion of the myocardium. This technique requires matter of time, which will bring technological
only small amounts of Gd-DTPA for each first- improvements allowing higher SNR, before
pass study. imaging of arteriosclerotic plaques in the coro-
The study is easily repeated to allow compari- nary arteries will become commonplace in a
son between baseline and stress conditions. This clinical setting.
technique is superior to radioisotope imaging Spatial Resolution.  Spatial resolution in radio-
because of better spatial resolution and better isotope imaging is approximately 2 lp/cm (5 mm)
contrast resolution, and it is also less susceptible because of the intrinsic limitations of the gamma
to artifact from soft tissue attenuation. ray detector. On the other hand, MRI can
During perfusion imaging, the Gd-DTPA routinely provide image resolution of 5 lp/cm
contrast is taken up much more slowly in in- (1 mm) with a 256 × 256 image matrix and
farcted tissue than in well-perfused myocardium. 25 cm FOV.
372 PART VI  Applications

Question: What is the pixel size and limiting For a typical late enchancement study per-
spatial frequency of a 256 × 256 image matrix formed on a 1.5 T MRI a phased-array coil is
and 10 cm FOV? used so that parallel imaging can be employed
100 mm and the scan can be acquired in an 8 to 12 second
Answer: Pixel Size =
256 breathhold. The wall segments are analyzed
= 0.4 mm for hyperenhancement from the trapping of Gd
−1
 0.8 mm  within the dead cardiac tissue.
Spatial Frequency =  
 1p  Often MRI can discriminate subendocardial
= 1.25 lp /mm infarcts, appearing on the inner surface of the
= 12.5 lp /cm heart, from transmural infarcts that go straight
through the cardiac wall. Nuclear medicine
Contrast Resolution.  Cardiac MRI with radio- studies, such as SPECT and PET, can reveal
isotopes has two deficiencies that restrict con- whether an infarct exists but these images do not
trast resolution. The count rate is low for photon have adequate spatial resolution to determine the
imaging because of the safety restrictions on infarcts’ extent or severity.
the amount of radioactive material that can Distinct late hyperenhancement patterns exist
be administered. Overlying tissue, especially in also in various nonischemic heart diseases and
obese patients, attenuates the already low signal. their visualization may ultimately aid diagnosis.
On the other hand, the intrinsic values of the For instance, consistent late hyperenhancement
MRI tissue characteristics—PD, T1, and T2— patterns have been observed in myocarditis,
have a far greater range of values than the detec- hypertrophic and dilated cardiomyopathies, and
tive quantum efficiency (DQE) of a scintillation systemic vasculitis.
crystal. Proper pulse sequence selection will
accentuate these ranges.
Magnetic Resonance Angiography
It is now possible to determine whether heart
tissue is dead or potentially recoverable using Fluoroscopically guided cine of the coronary
delayed imaging after injection of a Gd contrast arteries remains the gold standard for evaluat-
agent. This strategy takes advantage of the dif- ing coronary artery disease (CAD). However,
ference in signal intensity between normal and approximately 30% of coronary artery angio-
dead (infarcted) tissue. The presence of delayed grams are negative for CAD.
enhancement suggests infarcted myocardium. Because it is invasive, coronary artery angiog-
The difference in T1 between infarcted and raphy is accompanied by a finite risk of infection,
normal myocardium reaches a maximum 10 to infarction, and even death. MRA is a logical
20 minutes after the injection of Gd contrast. substitute because it is noninvasive and consider-
Most commonly, an inversion recovery- ably less expensive.
prepared fast spoiled gradient echo sequence is
used. The most important aspect of this sequence X-ray imaging of the coronary arteries has better
is the nonselective inversion recovery prepulse, spatial resolution than MRA.
for which the inversion time is estimated to null
myocardium and enhance T1 differences between Gradient echo TOF MRA with segmented
the normal myocardium in the infarcted tissue, k-space filling is successful in imaging coronary
which contains Gd. arteries down to perhaps 3 mm in diameter in a
A 180° pulse is used to invert MZ. The TI single breathhold. With such techniques, laminar
delay time is carefully selected to null the normal flow appears bright, and turbulence appears as a
myocardium, often using a special calibration signal void.
pulse sequence, and is typically set to a value Three-dimensional (3D) MRA has several
around 175 to 250 ms at 1.5 T. advantages over two-dimensional (2D) MRA. A
 CHAPTER 26  Cardiac Magnetic Resonance Imaging 373

3D signal acquisition has less operator error. transform (2DFT) MRA, or three-dimen­-
Images acquired in 3D mode have better spatial sional Fourier transform (3DFT) MRA?
resolution, better contrast resolution because 4. Which is superior for imaging the perfusion
of higher SNR, and extensive multiplanar and of the myocardium: radioisotope scans,
oblique planar image reconstruction. x-ray imaging, or MRI?
5. List the four MRI views required for com-
plete evaluation of the heart.
6. What minimum hardware requirements are
CHALLENGE QUESTIONS
advised for cardiac MRI?
1. List the four leading causes of death in the 7. Discuss the segmentation of k-space as it is
United States, from highest to lowest. used for cardiac MRI.
2. Describe the appearance of the coronary 8. How can one enhance the T2W of SE cardiac
arteries when imaged with GRE time of images?
flight MRA. 9. How is cardiac MRI speeded by the applica-
3. Which has best spatial resolution when im- tion of HFI?
aging contrast-enhanced coronary arteries: 10. Describe the proper positioning of ECG
x-ray angiography, two-dimensional Fourier leads during gated cardiac MRI.

"We're very happy with your work and want to reward you.
Funds though are tight so we've decided to name our new MRI
after you. However, you'll have to change your name to Ultravision."
PART
VII
Safety
 CHAPTER

27 
Contrast Agents and Magnetic
Resonance Imaging

OBJECTIVES
At the completion of this chapter, the student • Describe paramagnetism and how such an
should be able to do the following: agent enhances image contrast.
• List two general approaches to magnetic • Distinguish between positive and negative
resonance imaging (MRI) contrast contrast and describe how each is produced
enhancement. by reduced T1 and T2.
• Identify three routes for administration of • Identify the blood–brain barrier and how it
MRI contrast agents and give an example affects the use of MRI contrast agents.
of each.

OUTLINE
Approaches to Contrast Effect on Magnetic Blood–Brain Barrier
Enhancement Resonance Signal Intensity Integrity
Altering Hydrogen Content Other Possible Contrast Tissue Perfusion
Altering the Local Magnetic Agents Angiography
Field Clay Minerals Future Directions
Available Contrast Agents Iron
Contrast Agents Specific for Clinical Applications
Magnetic Resonance Tumor Localization and
Imaging Characterization

KEY TERMS
Blood–brain barrier Relaxation centers Superparamagnetic
Chelating Sensitivity
Paramagnetic Specificity

375
376 PART VII  Safety

With magnetic resonance imaging (MRI), the

naturally available contrast among tissues in
the body is good. It was originally assumed
that contrast enhancement (CE) would not
be needed in MRI. Even without the use of a
contrast agent, MRI is exceptionally sensitive
to detecting pathologic conditions. However,
as the field develops, it is becoming clear that
under some circumstances the addition of a
contrast agent greatly increases the diagnostic
value of MRI by improving disease sensitivity
and specificity and by delineating pathologic
processes.
In some areas of the body, adjacent tissues
have similar MRI appearance, and CE is needed
for adequate differentiation. CE is also useful
for delineating the structural integrity of the FIGURE 27-1  Biodistribution of a contrast agent in an
blood–brain barrier. The addition of a contrast astrocytoma. (Courtesy Michael Mawad, Houston, TX.)
agent allows imaging of areas of normal versus
decreased tissue perfusion. In some cases, pathol-
ogy can be better identified because of the differ-
ent speeds with which a contrast agent enters or not safe to use a sufficiently large change in a
washes out of tumors as compared with the sur- living organism to make this approach useful in
rounding normal tissue. vivo. However, it is possible to safely alter the
The properties needed in an MRI contrast chemical characteristics of living tissue suffi-
agent are the same as those needed for contrast ciently to change the MRI parameters and there-
agents in other imaging modalities. The agent fore image appearance.
should alter contrast at low concentrations, and In MRI, the image contrast is influenced by
its effect on contrast should be dose dependent. a number of factors, including the hydrogen
It must be nontoxic and nonreactive, as well as content (PD), spin-lattice relaxation (T1), spin-
rapidly cleared from the body. spin relaxation (T2), flow through the area
The biodistribution of a contrast agent is being examined, and the magnetic susceptibility
exceedingly important. An ideal contrast agent of the tissue. If any of these parameters are
distributes only to the organ, tissue, or pathology changed by the addition of a contrast agent, the
of interest (Figure 27-1). magnetic resonance (MR) image appearance is
also changed.
When the use of an MRI contrast agent
APPROCHES TO CONTRAST causes the tissue of interest to appear brighter,
ENHANCEMENT positive CE occurs. The CE is negative if the
The normal MRI appearance of a tissue depends tissue of interest is darker when a contrast agent
on both the physical characteristics of the tissue is used.
(such as viscosity and temperature) and the Although MRI contrast agents have tradition-
chemical characteristics that influence proton ally been used for positive contrast, it is becom-
density (PD), spin-lattice relaxation (T1), and ing more common to use them to reduce contrast,
spin-spin relaxation (T2). as in susceptibility-weighted cerebral perfusion
Although changing a physical property of a imaging. Box 27-1 presents one classification for
sample is useful for in vitro studies, it is generally MRI contrast agents.
 CHAPTER 27  Contrast Agents and Magnetic Resonance Imaging 377

variations in the local magnetic environment are

BOX 27-1 A Classification of Magnetic
averaged (rather like the blurring in vision that
Resonance Imaging Contrast
Agents by Route
a person experiences when spinning around
of Administration rapidly).
However, if the hydrogen ion is slowed, local
Via Ingestion magnetic field variations become perceptible.
• Soluble metal ion (ferric ammonium citrate) This causes a shortening of the relaxation times,
• Soluble metal ion chelates (gadolinium DTPA) particularly T2. This approach is limited but has
• Insoluble particulates (clay, barium, and iron oxide) been applied with some success to CE of the
Via Inhalation gastrointestinal tract.
• Oxygen Paramagnetic/Superparamagnetic Agents
Via Intravenous Injection
The most versatile approach to altering the local
• Metal ion chelates (gadolinium DTPA) magnetic environment of spins is to administer
• Nitroxide-stable free radical an agent that is paramagnetic or superparamag-
netic. The paramagnetic agents are most com-
monly based on gadolinium, dysprosium, or
manganese. The paramagnetic property is the
result of unpaired electron spins in certain
Altering Hydrogen Content
electron orbital shells of transitional metals or
The most obvious approach to CE is to alter lanthanides. For example, contrast is improved
the hydrogen content (PD) of tissue. Examples after administration of gadolinium-diethylene-
of this approach are water loading to increase triamine-pentaacetic acid dimeglumine (Gd-
the total signal from the kidneys and bladder DTPA) (Figure 27-2).
and administering diuretics to decrease hydra- Iron oxide particles are the superparamagnetic
tion of tissue. agents used. Such particles have high magnetic
An alternative approach is to remove the susceptibility and create a relatively large regional
source of the signal entirely by administering a gradient magnetic field. Such a gradient readily
hydrogen-free contrast agent that fills the area of influences water molecules diffusing close to the
interest. Another approach is to use oxygen. iron oxide particles. Reduced T1 and T2 result,
Unfortunately, although oxygen is paramagnetic, with the degree of CE dependent on many param-
the effect is too weak for clinical use. eters, including particle composition, size and
The amount of MR signal arising from tissue concentration, and pulse sequence used.
can also be changed by altering the ratio of water All currently approved clinical contrast agents
to fat in that tissue. The relaxation characteris- fall into this category. Paramagnetic and super-
tics of fat and water are quite different, so greatly paramagnetic agents are all metal ions or crystals
increasing the amount of fat can increase the that act like small bar magnets because they
available signal. This approach has been used for possess one or more electrons whose magnetic
imaging of the gut, with oral administration of field is not canceled by another electron of oppo-
fat-containing compounds such as mineral oil. site spin.
When such agents are placed in a magnetic
environment, such as when they are administered
Altering the Local Magnetic Field
as part of an MRI procedure, magnetization is
Motion Reduction.  One approach to changing induced within the contrast agent. A hydrogen
the magnetic environment perceived by spins is atom that is close experiences the magnetic field
to administer an agent that greatly slows the of the contrast agent in addition to the applied
normal motion of hydrogen. Normally, a hydro- static magnetic field (B0). This increase in per-
gen ion is tumbling quickly enough that any ceived magnetic field alters both T1 and T2.
378 PART VII  Safety

A B
FIGURE 27-2  A cyst in the right kidney is apparent in the unenhanced image (A) but more clearly demonstrated in the
90-second postcontrast image (B). (Courtesy Todd Frederick, Dallas, TX.)

FIGURE 27-3  Paramagnetic agents contain relaxation centers.

Therefore such nuclear species are referred to as shortening is more significant because relative T2
relaxation centers (Figure 27-3). values vary more than T1 values.
The use of CE is helpful for imaging many
MRI contrast agents reduce T1 and T2 relaxation areas of the body. Applications of CE magnetic
times. resonance angiography (MRA) are used for many
vessels of the body (Figure 27-4).
These contrast agents are not visualized An additional generality deals with radio­
directly by MRI, but rather, alter the MR signal frequency (RF) pulse sequence selection to
indirectly by shortening relaxation times. The optimize the CE. Paramagnetic contrast agents
magnitude of the effect on each relaxation time are most effective with T1-weighted (T1W) RF
varies with the paramagnetic or superparamag- pulse sequences and result in an increase in signal
netic species and its concentration in tissue. and therefore positive CE. Conversely, super-
In general, low concentrations affect T1 more paramagnetic contrast agents affect T2 more
than T2 because of intensified spin-lattice inter- than T1; therefore T2-weighted (T2W) RF pulse
action. At high concentration, the resulting T2 sequences are more appropriate. The result of the
 CHAPTER 27  Contrast Agents and Magnetic Resonance Imaging 379

A B

C
FIGURE 27-4  These images—A, subclavian contrast-enhancement (CE) magnetic resonance angiography (MRA) dem-
onstrating a right subclavian occlusion, B, pulmonary CE MRA, and C, renal CE MRA—demonstrate the versatility of CE
MRA. (Courtesy Todd Frederick, Dallas, TX.)

use of these contrast agents is a reduced signal are gadoterate meglumine (Dotarem, Guerbet
intensity—negative CE. Laboratories, Aulnay-sous-Bois, France), gado-
teridol (ProHance, Bristol-Myers Squibb Diag-
nostics, Princeton, NJ), and gadodiamide
AVAILABLE CONTRAST AGENTS (Omniscan, Sanofi-Winthrop Pharmaceuticals,
New York, NY). All of these contrast agents
Contrast Agents Specific for contain the paramagnetic metal ion gadolinium,
Magnetic Resonance Imaging which has seven unpaired electrons in its outer
In 1988 the first commercial contrast agent shells.
(Magnevist, Berlex Laboratories, Wayne, NJ) Chemical Structure and Toxicity.  A major
was clinically introduced. Magnevist is Gd-DTPA difficulty in the development of any of the para-
or gadopentetate dimeglumine. magnetic metals as contrast agents has been
Three additional contrast agents are specifi- to diminish the toxicity of these compounds to
cally approved for use in clinical MRI. These clinically acceptable levels. The LD50 (lethal
380 PART VII  Safety

H H available for clinical MRI are administered

O intravenously and distributed into the blood
O and extracellular space. Therefore they are con-
CH3
O N
sidered nonspecific agents in that they are not
H
O taken up by a particular organ, tissue, or lesion
Gd N
type.
O N All paramagnetic contrast agents clear from
O
the body rapidly after administration. The elimi-
O nation half-life is generally in the range of several
N hours.
H O xH2O
H3C

O Effect on Magnetic Resonance

FIGURE 27-5  Chemical structure of gadolinium-DTPA. Signal Intensity
These contrast agents reduce both T1 and T2
relaxation times. The relaxation time changes
dose for 50% of subjects) of gadolinium chloride cause a decrease in signal intensity on T2W and
in rats, for example, is less than 1 mM/kg after T2*W images and an increase in signal intensity
intravenous administration. The gadolinium is on T1W images.
still present in the body several days after In most clinical applications, T1W images are
administration. used for contrast-enhanced studies. Areas of
If these metals are allowed to interact with increased signal intensity indicate the presence of
the human body, they are quite toxic. A con­ the contrast agent.
centration of gadolinium as low as 1 mM/kg Sometimes a T2W or T2*W image is used, in
is sufficient to cause obvious neurotoxicity in which case the presence of the contrast agent
brain. is indicated by a decrease in signal intensity. If
It is necessary to bind agents to another sub- the concentration of contrast agent in an area
stance that decreases their interactions within becomes extremely high, it is possible for the
the body to decrease the immediate toxicity of effect on the T2 to dominate even the T1W
potential paramagnetic contrast agents and to image, causing a loss of signal intensity on both
increase the speed with which they are cleared types of images.
from the body. When gadolinium is bound to
the chelating agent DTPA (diethylenetriamine-
Gadolinium contrast agents shorten T1 and T2,
pantaacetic acid) (Figure 27-5), the LD50 in rats
but the principal image contrast results from
changes from 1 to 10 mM/kg, and the compound
shorter T1.
is eliminated from the body within hours.
The term chelating refers to the binding of one
substance with another. The acute toxicity of the CE is often equated with bright tissue. With
paramagnetic contrast agents approved for clini- the use of paramagnetic contrast agents, the
cal MRI is quite low. Few adverse reactions have desired effect, bright tissue, is the result of the
been reported. shortening of T1 relaxation time.
Therefore imaging techniques that emphasize
Gadolinium is chelated to DTPA to reduce T1 relaxation should be used to maximize the
toxicity. efficacy of the contrast agent. The pulse sequences
frequently used are short repetition time (TR),
Route of Administration and Biodistribution.  spin echo (SE) sequences, and inversion recovery
The paramagnetic contrast agents currently sequences.
 CHAPTER 27  Contrast Agents and Magnetic Resonance Imaging 381

A B
FIGURE 27-6  These images through the inferior aspect of the right lobe of the liver—A, FSE T2W and B, CSE T1W—
show three levels in the stomach, air, water, and a barium bentonite mixture. (Courtesy Michael Davis, Boston, MA.)

OTHER POSSIBLE CLINICAL APPLICATIONS

CONTRAST AGENTS
There are differences in the way contrast agents
are used for different pathologic conditions and
Clay Minerals
organs. Intravenous agents distribute throughout
The inert clay minerals kaolin and bentonite the body quite rapidly, and the total relaxation
have been used for clinical MRI of the gastroin- time shortening is greatest when the concentra-
testinal tract. Kaolin and bentonite are adminis- tion of the agent is greatest. Thus an image
tered orally in an aqueous suspension where they acquired soon after contrast administration
mix with the gastric contents and pass through shows the greatest change in signal intensity.
the gastrointestinal system (Figure 27-6). They However, acquiring the postcontrast MRI
are nontoxic and classified as safe for general use immediately after administration does not neces-
by the Food and Drug Administration (FDA). sarily give the maximum contrast between the
Kaolin is used routinely as an antidiarrheal pathology and surrounding tissue. The reason is
agent (Kaopectate). These substances shorten T1 that the contrast agent washes out of normal
and drastically shorten T2. At doses reasonable tissue faster than out of abnormal tissue. Thus in
for clinical use, T2 can be shortened sufficiently some applications, a delayed image is preferred.
to result in a signal void on both T1W and T2W
images.
Tumor Localization
and Characterization
Iron Contrast agent administration is used extensively
Iron in the form of ferric ammonium citrate to improve tumor localization and characteriza-
(Geritol), when administered orally, has been tion. The postcontrast administration appear-
used with some success for contrast-enhanced ance of the tumor depends on the tumor type,
imaging of the gastrointestinal tract. This is a tumor location, and the timing of the postcon-
paramagnetic substance that acts to decrease trast MRI. In most cases, the contrast agent is
relaxation times in the same manner as other used to increase the signal intensity of the tumor
paramagnetic contrast agents. Ferric ammonium on T1W images.
citrate has an extremely low toxicity and is cat- CE allows visualization of extremely small
egorized as quite safe. tumors and improves delineation of the tumor
382 PART VII  Safety

margin. Highly vascular tumors show the most Bolus injection with fast imaging allows for
CE. In breast and liver imaging especially, both digital subtraction angiography and multipass
early and delayed images may be useful, depend- imaging of the vessels. This method is particu-
ing on the suspected pathologic condition. In larly applicable to angiography of the brain
some cases, there is also evidence that CE may because the BBB keeps the contrast agent within
allow differentiation of benign from malignant the vasculature.
lesions. Normally, T1W images are used for angiogra-
phy, so presence of the contrast agent increases
the signal from the blood. This increases the
Blood–Brain Barrier Integrity
contrast between the vasculature and surround-
Normally, contrast agents cannot enter the ing tissue. A disadvantage of this technique is
central nervous system. They are unable to cross that arteries are not easily distinguished from
the blood–brain barrier (BBB) that protects the veins.
brain from substances in the blood. An excep-
tion is the pituitary gland, which does not have
FUTURE DIRECTIONS
a BBB.
However, many types of pathologic conditions Many new approaches to CE for MRI are being
disrupt the BBB. Contrast agents can enter brain actively explored. It is expected that in the future,
tissue in areas in which the BBB is disturbed, radiologists will have a variety of different
resulting in increased visibility of many types of contrast agents available, many of them much
brain abnormalities. In some cases, lesions that more specific in their effects than those presently
cannot be seen otherwise become visible. Visual- approved for clinical use.
ization of small lesions, such as early metastases, Several paramagnetic contrast agent candi-
may require delayed imaging or an increased dates in which the biodistribution is altered are
dose of contrast agent. under development. For example, binding a
paramagnetic contrast agent to microspheres or
albumin is being used to create contrast agents
Tissue Perfusion that stay within the blood. Also, binding to
An exciting new application of contrast agents in monoclonal antibodies is being attempted in an
MRI is the combining of contrast agent admin- effort to target specific diseases. Binding to lipo-
istration with a very fast image acquisition. This somes of different types has potential for target-
allows dynamic studies to be performed that can ing particular organs.
aid in assessing organ and tissue function. In addition, paramagnetic contrast agents for-
Contrast-enhanced perfusion sensitive imaging mulated for oral rather than intravenous admin-
has the potential to allow evaluation of regional istration are being tested for gastrointestinal
blood flow, blood volume, and tissue perfusion. imaging.
These techniques have already been used to Several contrast agent candidates are based
delineate areas of decreased or absent blood flow on iron oxide particles. Agents for both oral
in the brain and heart. (gastrointestinal imaging) and intravenous (blood
pool and reticuloendothelial system imaging)
administration are being tested. Work is also
Angiography under way to make agents that target particular
Although most magnetic resonance angiography receptors or pathologic conditions.
(MRA) is presently performed with special image A fluorocarbon contrast agent candidate
acquisition pulse sequences, it is also possible to (perfluorohexybromide) is being developed that
obtain angiograms by combining the informa- evaporates at body temperature. When it is
tion from normal and contrast-enhanced images. administered orally, it forms a gas in the gut, thus
 CHAPTER 27  Contrast Agents and Magnetic Resonance Imaging 383

producing a signal void. The gas is radiopaque, 4. What is the general appearance of tissue
has a faster transit time than barium, and appears after MRI contrast-enhanced imaging?
inert and nontoxic on initial testing. 5. What is the role of the BBB in contrast-
These developments indicate that with clever enhanced brain imaging?
applications of chemistry, the role of contrast 6. What is the principal MRI contrast agent?
agents in MRI will only increase in the future, as 7. Do superparamagnetic and paramagnetic
new compounds are developed to improve image contrast agents produce positive or negative
quality for a variety of diagnostic roles. contrast?
8. As a general rule, what is the effect of MRI
contrast agents on PD, T1 relaxation, and
CHALLENGE QUESTIONS
T2 relaxation?
1. What are some desired characteristics of any 9. The DTPA in gadolinium-DTPA is a chelat-
medical imaging contrast agent? ing agent. What is that?
2. Why should iron oxide particles make a 10. Why is chelation important? Why not ad-
good MRI contrast agent? minister the gadolinium in atomic or ionic
3. How are contrast agents classified according form?
to their route of ingestion?

"I don't understand all the mechanics but surely

we can make this work!"
 CHAPTER

28 
Magnetic Resonance
Imaging Artifacts

OBJECTIVES
At the completion of this chapter, the student • Recognize image aliasing, chemical shift, and
should be able to do the following: motion artifacts.
• Identify the three principal causes of magnetic • Describe the partial volume artifact and how
resonance imaging (MRI) artifacts. to reduce it.
• Define the difference between patient-related
and system-related artifacts.

OUTLINE
Magnetic and Radiofrequency Reconstruction Artifacts Noise-Induced Artifacts
Field Distortion Artifacts Patient Related Patient Related
Patient Related System Related System Related
System Related

KEY TERMS
Ferromagnetic material Misregistration artifacts Zipper artifact
Image artifact Off-resonance artifact
Magnetic susceptibility Truncation

An image artifact is a pattern or structure in the be interpreted easily and do not interfere with
image caused by a signal distortion related to the diagnosis. However, the addition of each new
technique of producing an image. Therefore it is imaging technique or pulse sequence brings the
an unwanted pattern or structure that does not possibility of new artifacts.
represent the actual anatomy. Image artifacts can The MRI artifacts illustrated here fall into
be misleading and interfere with diagnosis. several categories. The outline in Box 28-1 adopts
The complexity of magnetic resonance imaging a scheme of three principal classes of MRI arti-
(MRI) has unfortunately brought with it many facts, each of which can be identified as patient
confusing imaging artifacts. Luckily, many can related or system related. None of these artifacts

384
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 385

homogeneity of ±1 ppm (parts per million) over

BOX 28-1 Classification of Magnetic
a 20 cm spherical volume can be achieved by
Resonance Imaging Artifacts
many current MRI systems.
I. Magnetic and RF Field Distortion Artifacts
A. Patient related
A homogeneity of ±1 ppm per 20 cm volume
1. Foreign materials
means that a 1 T magnetic field does not vary by
2. Magnetic susceptibility
3. Body shape, conductivity, and extension more than ±1 µT throughout a 20 cm diameter
4. Chemical shift sphere, centered at magnet isocenter.
B. System related
1. Primary static magnetic field inhomogeneity
Distortion of the magnetic field can be caused
2. Gradient magnetic field inhomogeneity
by the patient or problems with the MRI system.
3. RF coil inhomogeneity
4. Gradient coil switching/timing inaccuracy A poorly shimmed magnet, gradient magnetic
II. Reconstruction Artifacts field miscalibration, chemical shift of the Larmor
A. Patient related frequency, tissue magnetic susceptibility, and
1. Aliasing (wraparound) foreign materials within or on the patient can all
2. Partial volume averaging result in distortion of the magnetic field.
B. System related
1. Truncation
2. Zero frequency Patient Related
III. Noise-Induced Artifacts Foreign Materials.  Metals have different effects
A. Patient related
on the local magnetic field depending on their
1. Voluntary motion
shape and location. Any metal is a potential con-
2. Involuntary motion
a. Bowel peristalsis ductor of electricity, which can support current
b. Respiration induced by the pulsing of the gradient and RF
c. Cardiac and vessel pulsation magnetic fields. The currents produced by gradi-
d. Swallowing ent coils will change the local magnetic field,
3. Fluid motion leading to image distortions. The currents pro-
a. Blood flow duced by the RF coils can lead to tissue heat and
b. Cerebrospinal fluid flow patient burns.
c. Urine movement Ferromagnetic materials, in particular, should
4. Misregistration be avoided since they can establish magnetic
B. System related
fields that persist independent of the pulsing
1. Frequency line/point
MRI secondary magnetic fields. Ferromagnetic
2. Phase line/star
3. Extraneous RF materials can produce not only a local signal
4. Off-resonance turning loss but also a warping distortion of the sur-
rounding areas. An example of such field distor-
tion is provided by the friendly “cone-head” seen
are characteristic of a particular manufacturer or in Figure 28-1.
model of MRI system. The typical metal material artifact has a
partial or complete loss of signal at the site of
the metal. Such metal can distort the local mag-
MAGNETIC AND netic field sufficiently so that the Larmor fre-
RADIOFREQUENCY FIELD quency for local spins is outside the frequency
DISTORTION ARTIFACTS range of the imaging system. Furthermore, metal
MRI requires an exceptionally uniform, homo- contains no hydrogen; the result is signal void at
geneous, primary magnetic field. Magnetic field that location.
386 PART VII  Safety

FIGURE 28-1  A small band of metal in this patient’s ponytail produced this “cone-head” artifact. Although the region
of interest was not affected by the artifact, the images were incinerated to avoid publicity about unanticipated adverse
bioeffects of magnetic resonance imaging.

Sometimes a partial rim of high signal intensity metal-induced change in the local magnetic field.
may be seen at the periphery of the signal void, The magnetic field lines are distorted, resulting
which allows differentiation of metal from other in a change in the local Larmor frequency (Figures
causes of focal signal loss. Usually, the degree of 28-3 to 28-5).
anatomical information loss due to such an arti- More mundane but relatively common metal-
fact is less than that seen on computed tomogra- lic materials that may be encountered include
phy (CT) images, because the loss is local and not belt buckles, keys, mascara and other makeup,
streaked across the image as in CT. foreign bodies in the eye or elsewhere, and even
However, ferromagnetic material artifacts can certain types of nylon found in clothing, such
actually cause very subtle, yet significant arti- as gym shorts and warm-up suits (Figures 28-6
facts. The classic example is orthodontic braces, to 28-9).
which cause severe signal dropout from the Magnetic Susceptibility Artifacts.  Magnetic
mouth and jaw. However, they can also cause susceptibility is a dimensionless quantity that
artifacts in the remote regions in the brain by indicates the relative degree of magnetization
distorting the slice selection process. occuring in a material that is placed in a mag-
netic field. Materials with magnetic properties
Distortion of the magnetic field results in a sus- that are different from the tissues in which
ceptibility artifact. they are embedded can cause a “susceptibility
artifact.”
Exceptions to this are usually caused by the Also there are differences in magnetic suscep-
screws in some metallic orthopedic devices, but tibility in the three major subtances normally
occasionally distortion is produced by small found within the human body: air (gas), bone,
metallic objects such as buttons, snaps, zippers, and soft tissue (water). Most soft tissues are
or barrettes (Figure 28-2). The distortion of diamagnetic and will change the applied, B0
the surrounding tissue is the result of the Text continued on p. 392
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 387

FIGURE 28-2  The metallic zipper in this patient’s shirt produced the noticeable loss of signal, as seen on this image.
The high signal rim at the edges of the void is typical.

FIGURE 28-3  An electrocardiographic lead makes a small, localized distortion of the anterior chest wall.
388 PART VII  Safety

FIGURE 28-4  A metallic stent, which was temporarily placed in this patient’s ureter, caused image degradation along
the entire path from the kidney to the bladder.
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 389

B
FIGURE 28-5  Multiple surgical clips and a Gianturco immobilization coil were in place at the time of this examination.
A, The immobilization coil causes a relatively large signal void within the liver. B, Some clips show only small amounts
of signal dropout, as seen in the lower abdomen of the same patient.
390 PART VII  Safety

B
FIGURE 28-6  Gradient echo imaging causes accentuation of local magnetic field inhomogeneities produced by ferro-
magnetic materials. A, The large area of signal dropout results from two small metallic bra clips that were pulled up
between this patient’s scapulae. B, A spin echo image obtained through the same level shows a more typical and less
prominent metallic artifact. A wraparound aliasing artifact of the arms onto the abdomen is also present.
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 391

B
FIGURE 28-7  A, A computed tomography image of the abdomen that is nondiagnostic because of the artifacts pro-
duced by surgical clips in the postoperative bed. B, The magnetic resonance image of this same patient reveals the
recurrent pheochromocytoma. A small signal void immediately above the mass resulted from one of the clips.
392 PART VII  Safety

TABLE 28-1 Magnetic Susceptibility

Values of Some Materials
Found in MR Images

Susceptibility
Material (ppm)
Gold −34
Pyrex glass −13.9
Water −9.1
Polystyrene −7.5 to −8.2
Whole blood (deoxygenated) −7.9
Human soft tissues −7 to −11
Bone marrow −3.5
Bone −0.9
Whole blood (oxygenated) −0.7
Air (NTP) +0.36
FIGURE 28-8  Metallic orthopedic devices can produce Aluminum +20.7
severe distortion and signal loss. The screws in the Har- Calcium +21.7
rington rods caused this image to be nondiagnostic. Titanium +182
Gadolinium solution (0.1 mol L−1) +249

In general, materials exhibit one of three

classes of magnetic susceptibility. Ferromagnetic
materials (iron, nickel, etc.), which were dis-
cussed in the previous section, strongly influence
the magnetic field and lead to major distortions
of all MR images. Paramagnetic and diamagnetic
susceptibilities are much weaker and are differ-
entiated by whether they increase or decrease the
local magnetic fields compared to B0. Aluminum
and calcium are examples of paramagnetic mate-
rials while water and most organic compounds
are classified as diamagnetic.
Materials with magnetic susceptibilities on the
FIGURE 28-9  Ferrometals may cause significant artifacts, order of only a few parts per million will distort
such as distortion of the orbital globes produced by eye
shadow.
the fields produced by the linear magnetic field
gradients, and can produce characteristic suscep-
tibility artifacts. These artifacts appear as adja-
magnetic field by −7 ppm to −10 ppm. Table cent bright areas (misregistered signals) and dark
28-1 lists the magnetic susceptibilities of several areas (no signal) in or near the magnetic material
tissues and materials which might be found in being imaged.
medical implants. Even a susceptibility difference One situation in which susceptibility artifacts
of less than 1 ppm can cause a measureable effect are problematic is the temporal region of the
on images produced with pulse sequences that head, near the sphenoid air sinuses. The tempo-
are most sensitive to susceptibility differences ral bones are next to the temporal air sinuses,
(i.e., fast gradient echo and echo-planar imaging). causing a magnetic field gradient that produces
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 393

a susceptibility artifact that can obliterate part of also become important in head images and pedi-
the temporal lobe of the brain in the image. atric body images produced by the emerging 7 T
Magnetic field variations at air-bone interfaces MRI systems.
produce some of the most serious susceptibility Chemical Shift.  Chemical shift artifacts are
artifacts in gradient echo images and echo-planar present wherever contiguous tissues have con­
images. Susceptibility artifacts can be minimized siderably different molecular organization. The
by using pulse sequences with very short TE artifact is seen as a bright rim of signal at one
values and/or 180° refocusing RF pulses. interface and a dark rim on the opposite side of
Body Shape, Conductivity, and Extension.  the particular organ, oriented in the frequency-
The patient’s shape, electrical conductivity, and encoded direction. The most prominent exam-
degree of coupling to the radiofrequency (RF) ples seen are at interfaces of tissues that have a
coil all become factors in creating inhomogeneity high fat content and other body tissues that are
of both the primary static magnetic field and the water-like.
transmitted RF pulse. Extension of a body part The molecular environment within fat causes
outside the area of maximum field homogeneity hydrogen nuclei in fat molecules to precess at a
will frequently cause a metallic-like artifact at the slightly lower frequency than the hydrogen nuclei
edges of this area. This curvilinear artifact con- in the water molecules of other soft tissues. The
forms to the shape of the magnetic field at the hydrogen electron of water is bound more tightly
edges and may have a characteristic pattern for to oxygen than that of fat is bound to carbon.
an individual magnet system. This hydrogen electron provides a magnetic
This situation can occur when imaging the shield to the proton spin that is stronger for the
shoulder in a tubular superconducting magnet. carbon-bound hydrogen than for the oxygen-
One of the advantages of open magnet configura- bound hydrogen.
tions is that the shoulder region can be moved to
magnet isocenter, thereby avoiding signal losses The chemical shift between fat and water protons
and misregistration issues. is about 3.5 ppm.
At B0 magnetic fields of 1.5 T or less the wave-
length of the radiofrequency excitation used to The result is a slightly higher Larmor fre-
excite the spins is on the order of a meter. As quency, 3.5 ppm, for soft tissue water spins
higher field strengths are used, for example at compared with fat spins. For example, at 1 T
3 T, the wavelength of the RF applied to the the 3.5 ppm difference results in a frequency
patient becomes smaller than the dimensions of difference of 149 Hz. At 1.5 T the frequency
the body. The dielectric properties of the tissues difference is 224 Hz, resulting in even more
can cause constructive and destructive interfer- dephasing.
ence patterns to occur, leading to a variation in Two components within a voxel with slightly
signal intensity. different resonance frequencies provide a hyper-
At 3 T, this can make the image intensity at intense signal if they have the same phase. They
the center of the body as much as 50% greater will cause a hypointense signal because of signal
than it is at the edges, producing a high degree cancellation if the magnetization of these two
of image intensity nonuniformity. The dielectric components is of opposed phase.
effect is dampened as the conductivity of the part Therefore fat will be slightly displaced because
under investigation increases. spatial localization is based on the hydrogen spin
Dielectric effects can only be mitigated by an frequency of water. The artifact is seen only in
advanced RF pulsing scheme using RF transmit the frequency encoding direction at interfaces
array coils, which is commonly referred to as where the bright signal of fat is shifted into or
“RF shimming.” While dielectric artifacts are away from the signal of the darker, water-based
important only in adult body images at 3 T, they soft tissues.
394 PART VII  Safety

FIGURE 28-10  The dark rim of signal along the right bladder wall occurs because of a chemical shift artifact at the
interface between pelvic fat and bladder urine and wall. The left side of the bladder is brighter, which could occasionally
be mistaken for edema or tumor infiltration of the normally darker bladder wall. The bright signal within the posterior
urine-filled bladder is an entry slice flow phenomenon more commonly seen in blood vessels. (Courtesy Wlad Sobol,
Birmingham, AL.)

The number of pixels shifted by the chemical echo time (TE) can be set so that the fat and water
shift artifact increases with magnetic field signals are imaged either in phase or out of phase.
strength. The number of pixels shifted may be When they are imaged in phase, maximum signal
decreased by acquiring the signal with a large is achieved. However, when they are imaged out
receiver bandwidth, however this comes with of phase, the signal from fat within a voxel par-
some loss in SNR. Therefore in systems with tially cancels the signal from water and creates a
1.5 T magnets and higher, stronger gradient boundary artifact around many organs.
magnetic fields are required so that imaging can
be done with larger bandwidths to reduce the
System Related
visibility of this artifact.
There are also many system-related causes
It is the receiver bandwidth and the B0 field of magnetic and RF field inhomogeneity. The
strength that control the prominence of the primary coil and each secondary coil may con-
chemical shift artifact. tribute to such artifacts.

The chemical shift artifact is most prominent Zipper artifacts can occur at any position along
at the border of the kidneys and perirenal fat the frequency- or phase-encoding gradients.
around the bladder and at the interface of retro-
orbital fat with the optic nerve and muscles The primary static magnetic field can never
(Figure 28-10). This shows up in the frequency- maintain perfect stability and may vary region-
encoding direction on conventional images and ally from day to day. In a similar manner,
in the phase-encoding direction in echo-planar each of the hardware components used to trans-
imaging (EPI). mit and receive RF signals and manipulate the
A second type of chemical shift artifact mostly gradient magnetic fields can do so with limited
shows up on gradient echo (GRE) images. The consistency. The images produced by surface
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 395

collagenous structures, even on short TE

sequences. However, at a precise angle, 54.7
degrees between the collagen fibers and the long
axis of the magnetic field (B0), the dipolar inter-
action falls to zero. This causes the T2 relaxation
time to lengthen to 22 ms, a 100-fold increase.
This effect produces hyperintensity on MR
sequences using a short TE and can be mistaken
for tendinopathy or other musculoskeletal
pathology.
The magic angle artifact is seen in tendons,
ligaments, menisci, articular cartilage, and
peripheral nerves. To correct for the magic angle
artifact, longer TE times can be used. Lengthen-
ing the TE causes regression of signal intensity
induced by the magic angle. The magic angle
effect has virtually no effect on T1 relaxation
FIGURE 28-11  The top of the head overlaps the lower time; therefore, T1-weighted sequences are not
neck in this image obtained with a circumferential neck affected.
coil. When a portion of the body is outside the coil, it can
wrap around the bottom of the image.
RECONSTRUCTION ARTIFACTS
Frequently observed MRI artifacts relate to RF
coils and GRE magnetic imaging techniques pulse sequence techniques used and computer
are especially sensitive to magnetic field reconstruction algorithms. They include alias-
inhomogeneities. ing, partial volume averaging, truncation, and
Zipper artifacts parallel to the frequency- quadrature artifacts.
encoding axis can be caused by residual trans-
verse magnetization, stimulated echoes, and
Patient Related
therefore, incomplete spoiling. Zipper artifacts
parallel to the phase-encoding gradient can be Aliasing.  The aliasing artifact is one of the most
caused by multiple sources of RF leakage and commonly encountered of this group. It was
feed through. mentioned briefly in Chapter 22. Aliasing, or
The magic angle artifact is seen on sequences “wraparound,” occurs when portions of the
with a short TE, such as T1W, PE, and gradient patient’s body are outside the prescribed field of
echo. It plays a role particularly in MSK imaging. view (FOV) but within the area of RF excitation.
The magic angle is a manifestation of the aniso- Therefore RF signals that cannot be properly
tropic effects of collagen in MR. The highly interpreted are produced.
organized molecular structure of collagen When hydrogen nuclei outside the area of
restricts the mobility of nearby water protons, interest are excited, the signal they return is inter-
thereby promoting dipole-dipole interactions preted to have originated from within the imaging
between them. FOV. It is then projected over the real portion
Normally, dipolar interactions between of the image on the opposite side of its actual
restricted water protons in collagenous struc- location.
tures cause very rapid T2 relaxation, on the
order of 250 µs. This rapid decay is responsible Aliasing results in a wraparound artifact.
for the low signal intensity seen in normal
396 PART VII  Safety

FIGURE 28-12  This illustrates the meaning of the term truncation.

The aliasing artifact occurs because the phase adequacy of the gradient magnetic coils or of
angles of the external nuclei are essentially equal pulse crafting may not be equal to the task of
to the nuclei in the imaging volume but on the precisely defining such thin slices.
opposite side of the image. The reconstruction
algorithm dutifully places these signals where
System Related
they should be (see Figure 28-6). This “wrap-
around” artifact is always in the phase-encoding Truncation.  Truncation is a geometric term that
direction; however, manufacturers have solved describes the cutting off or lopping of the vertex
this problem with the use of reconstruction of a cone or cylinder (Figure 28-12). The remain-
filters. An example is seen in Figure 28-11, in ing part is the truncated part.
which the top of the head (which was outside The truncation or “ringing” artifact seen in
the RF coil) projects over the upper thorax and Figure 28-13 appears as multiple, well-defined
lower neck. curved lines regularly conforming to the ana-
Partial Volume Averaging.  This artifact is tomic boundary. The truncation artifact is more
similar to that engaged in CT imaging. Partial pronounced when the number of phase-encoding
volume averaging results whenever the particular acquisitions is small.
structure of interest is contained within two con- Truncation occurs in areas where there is a
tiguous slices. The artifact is worse with thick great difference in signal intensity, such as inter-
slices and large voxels. faces of fat and air or fat and cortical bone. The
sharp contrast boundaries of these interfaces
The use of thin slices reduces the partial volume consist of high spatial frequencies. With a small
artifact. matrix, there are not enough data to represent
such high frequencies accurately. The orientation
However, thin-slice imaging requires more of the artifact can be along both frequency- and
time for signal averaging because it must attain phase-encoding axes.
an adequate signal-to-noise ratio (SNR), and Truncation artifacts can be particularly both-
more slices are necessary. Furthermore, the ersome in spinal imaging, where the cord can be
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 397

The zero-frequency artifact is often managed

by acquiring every other line of data with a 180°
phase shift. This does not eliminate the artifact,
but shifts it to edge of the image FOV, where it
is less troublesome.
A similar-appearing line artifact can be pro-
duced by RF noise from extraneous sources and
is sometimes called a free induction decay (FID)
line (Figure 28-15). The source of such RF emis-
sions can be any electrical appliance or broadcast
operating at the resonance frequency of the
magnet and authorized by the Federal Commu-
nications Commission. This artifact is not nor-
mally located on the centerline.

NOISE-INDUCED ARTIFACTS

Patient Related
Motion.  An additional group of artifacts is
FIGURE 28-13  This truncation artifact has multiple, caused by voluntary, involuntary, and even
evenly spaced, curvilinear lines of bright and dark signal microscopic physiologic motion. As with pho-
that follow the contour of the back of the head. When tography, a picture of a moving object cannot
the lines project in both the phase- and frequency- easily be taken.
encoding directions, the artifact is not the result of motion.
A magnetic resonance (MR) image is severely
degraded if motion occurs during the imaging
time. This is especially true if the motion occurs
near the middle of the acquisition. Physiologic
motion of the blood, heart, larynx (swallowing),
distorted or a false syrinx produced. Truncation diaphragm, bowel, and cerebrospinal fluid (CSF)
artifacts can be reduced by increasing the matrix in the brain and spinal canal can cause various
size or reducing the FOV. types of motion artifacts.
Zero Frequency Artifact.  A zero line or zipper The most prominent motion artifacts are
artifact is caused by RF feed-through from the poorly defined regions having a smeared appear-
RF transmitter along the frequency-encoding ance of the image in the area of motion. These
direction at the central or reference frequency of are called “ghosts.” Repetitive motion of a linear
the imaging sequence. It is typically seen on or curvilinear surface, such as the diaphragm or
images in which only one average is acquired heart, can produce this typical “ghosting” or
because no signal is obtained in the centerline of irregular wavelike lines of increased and decreased
the matrix. signal.
The result is a segmented line extending across The artifact appears in the phase-encoding
the middle of the FOV in the frequency-encoding gradient regardless of the direction of the motion
direction and having a zipperlike appearance because there is much more time in the phase-
(Figure 28-14, A). Occasionally, a single point, encoding direction (hundreds of milliseconds)
either bright (Figure 28-14, B) or dark (Figure than in the frequency-encoding read direction
28-14, C), is the only manifestation of this (tens of microseconds) for motion to make its
artifact. effects apparent.
398 PART VII  Safety

B C
FIGURE 28-14  A, The “zipper” artifact is always at the center of the field of view but not necessarily through the
center of the patient. B, A central point of brightness, as shown over the right midbrain, could potentially cause a mis-
diagnosis. C, The central point artifact can also be dark, as seen on this second echo image.
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 399

changed with most systems, rendering the arti-

fact less bothersome.
The CSF flow in the region of the foramen of
Monro and Magendie can produce bright foci in
the region of the midbrain and brain stem. This
signal occurs in the phase-encoding direction and
may mimic various lesions such as infarcts, mul-
tiple sclerosis plaques, and gliosis.
The flow of CSF in the cervical thecal sac
can also produce high signal artifacts overlying
the cord, which can obscure this area or be mis-
taken for pathologic conditions. This artifact is
relatively easy to recognize because it is more
linear and parallel with the spine on sagittal
images, projecting in the phase-encoding direc-
tion (Figure 28-18).
Blood flow artifacts can be extremely useful
because most vessels are seen as black on
FIGURE 28-15  Lines produced by random radiofre- conventional spin echo (SE) pulse sequences.
quency (RF) noise can occur whenever RF shielding of the The blood returns a relatively strong signal on
room is inadequate. Extraneous RF noise at two frequen-
proton density–weighted (PDW) and T2-weighted
cies shows up on this image.
(T2W) images if it is stationary or flowing
slowly (Figure 28-19), but the artifact created
by its motion produces the typical signal void
instead.
Ghost artifacts appear in the phase-encoding A problem can occur when flowing blood is
direction when a tissue moves at a rate much oriented diagonally within the imaging plane. In
faster than the image signal acquistion time. this case, the nuclei remain in the imaging FOV
so that the emitted RF signal is received even
Techniques for decreasing motion artifacts though there may be rapid flow. The resultant
without increasing imaging time have been devel- displaced blood signal parallels the dark blood
oped for abdominal imaging. These include vessel and appears as an adjacent bright line
various corrective algorithms, physical restric- (Figure 28-20).
tion of the motion of the anterior abdomen, and Misregistration.  If patient motion occurs
rapid imaging techniques that can be performed between a mask image and a subsequent image,
during breathholding (Figure 28-16). Use of the subtracted image contains misregistration
short repetition time (TR) and TE with multiple artifacts. The same anatomy is not registered in
acquisitions produces good images by averaging the same pixel of the image matrix.
the motion artifacts (Figure 28-17). This type of artifact can frequently be
The problems with artifacts produced by eliminated by reregistration of the mask, that
swallowing and bowel peristalsis are less easily is, by shifting the mask by one or more pixels
handled but are generally not as severe as those so that superimposition of images is again
caused by vascular and respiratory motion. Fast obtained. However, reregistration can be a
imaging techniques that are based principally on tedious process.
GREs appear to provide a partial, if not com- Often when one area of an image is reregis-
plete, solution to such artifacts. Furthermore, the tered, another area will become misregistered.
direction of the phase-encoding gradient can be This can be controlled on some systems by region
400 PART VII  Safety

B
FIGURE 28-16  A, This coronal image of the abdomen was obtained during breathholding. Notice the complete absence
of blurring that is usually seen near the diaphragm. B, A similar image obtained in an equal amount of time during quiet
respiration shows poor definition of the subdiaphragmatic structures.
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 401

of interest (ROI) reregistration. Many systems

can reregister not only in increments of pixel
widths but also down to 110 pixel width.

System Related
The off-resonance artifact is simply the degrada-
tion of the image as a result of inexact tuning of
the RF transmitter and/or receiver to the Larmor
A frequency, resulting in overall noisy images. This
mainly occurs when imaging small structures off
the magnet isocenter, such as elbows or hands.
A “bleeding” artifact results when inexact
duplication of the RF pulses occurs or the gradi-
ent magnetic fields between each acquisition are
incorrectly formed (Figure 28-21). This produces
slight inaccuracies in the placement of signals
into the appropriate pixel; the result is a smeared
image of wavelike patterns similar to ghosting
from motion. The off-resonance artifact can also
B appear similar to the truncation artifact, but not
as regular and well-defined.
FIGURE 28-17  A, The motion artifact can be masked
because it is, for the most part, a random signal that
subtracts out in multiple, average images. A comparison
of the same patient with only two averages shows a con-
siderable amount of distortion from respiratory motion.

FIGURE 28-18  The loss of cerebrospinal fluid signal at and immediately below the aqueduct and foramen of Magendie
mostly results from the rapid dephasing of spins caused by turbulent flow.
402 PART VII  Safety

A B

C D
FIGURE 28-19  These four contiguous images, from (A) superior to (D) inferior, show the normal, black signal void in
the inferior vena cava and right iliac vein resulting from rapidly flowing blood. The left iliac vein is compressed as it passes
beneath the left iliac artery. It has a relatively bright, intraluminal signal in the more slowly flowing portion, which is distal
to the point of compression.
 CHAPTER 28  Magnetic Resonance Imaging Artifacts 403

B
FIGURE 28-20  A, A petrosal vein is seen traveling obliquely through this transaxial image of the brain. This demonstrates
a normal signal void. B, The second, even echo of this four-echo sequence shows a double linear track at the site of the
vein. Even echo rephasing of the blood signal has occurred, accounting for the bright line, but it has been displaced
laterally.
404 PART VII  Safety

CHALLENGE QUESTIONS
1. Define image artifact.
2. What is an off-resonance artifact, and how
does it appear?
3. How can a partial volume artifact be
reduced?
4. Given the data in Table 28-1, would you
expect a magnetic susceptibility artifact
between bone and bone marrow? How
about oxygenated and deoxygented blood?
5. What is a misregistration artifact?
6. What causes the chemical shift artifact?
How can it be minimized?
7. What is a ghost artifact? In which direction
FIGURE 28-21  The “bleeding” artifact can be mistaken will it appear?
for a displacement of signal as a result of motion; however, 8. Describe the cause and appearance of a
this artifact is a more diffuse smearing of signal, both zipper artifact.
within and outside of the body. 9. What is a wraparound artifact? How can it
be avoided?
10. Define the term truncation.
 CHAPTER

29 
Biological Effects of
Magnetic Resonance
Imaging

OBJECTIVES
At the completion of this chapter, the student • Define SAR and identify recommended
should be able to do the following: exposure limits.
• Differentiate between stochastic and • Identify the cause of the banging noise
deterministic effects. in MRI.
• Identify the three magnetic resonance imaging • Discuss hazards imposed by ferromagnetic
(MRI) energy fields potentially capable of projectiles.
producing a harmful response.
• Describe the possible human responses to each
of the three MRI energy fields.

OUTLINE
Magnetic Resonance Imaging Human Responses to Magnetic General Safety Considerations
Energy Fields Resonance Imaging Patient Evaluation
Static Magnetic Fields Effects of Static Magnetic Ferromagnetic Projectiles
Transient Magnetic Fields Fields Pregnancy and Magnetic
Radiofrequency Fields Effects of Transient Resonance Imaging
Combined Magnetic Magnetic Fields
Resonance Imaging Fields Effects of Radiofrequency
Low-Frequency Fields
Electromagnetic Fields Recommended Guidelines

KEY TERMS
Intensity-response relationship Polarized Threshold intensity
Nonionizing Radiofrequency field Transient magnetic field

405
406 PART VII  Safety

Many noninvasive medical imaging modalities relationship is known as a threshold, nonlinear

are currently available, such as x-ray imaging, relationship.
radioisotope imaging, ultrasonography, and It is inappropriate to use the term dose when
magnetic resonance imaging (MRI). Being non- describing the energy fields associated with MRI.
invasive, each of these modalities is inherently Intensity is a better term, because it implies
safe. Ionizing radiation can induce malignant bathing a patient in an agent, rather than admin-
disease and genetic mutations, although the risk istering a certain quantity of the agent to the
of such responses is extremely low. patient.
Regardless of the type of MRI energy field
Ultrasonography and MRI prevail in safety considered, there is a level of intensity, the thresh-
considerations because neither uses ionizing old intensity (IT), below which no response is
radiation. elicited. Below IT, MRI is entirely safe. Above IT,
the response to MRI exposure first increases
Human response to ionizing radiation follows slowly and then more rapidly until 100%
a linear, nonthreshold, dose-response relation- response is observed. A response that exhibits a
ship (Figure 29-1). Nonthreshold means that no threshold and a dose-related severity is a deter-
dose of ionizing radiation is considered to be ministic response.
absolutely safe. Even the smallest dose carries a
risk of response, although such risk may be lower Responses to MRI are deterministic.
than that from other everyday risks.
At low radiation doses, the increase in response Because MRI is still developing, all of the
is exceedingly low. Only after rather high radia- radiobiological questions are not completely
tion doses (e.g., 0.25 Gy or greater) would the answered. It does appear that MRI, as it is cur-
anticipated response be detectable. rently used, is safe for all workers and patients.
Such responses to ionizing radiation are sto- Experimental observations suggest that IT for
chastic. There is no threshold dose, and the all of the MRI energy fields is considerably
response is an increase in incidence, such as higher than the intensities used for clinical
cancer or leukemia, rather than severity. examinations.
Exposure to the energy fields of MRI,
both individually and collectively, results in a
MAGNETIC RESONANCE IMAGING
fundamentally different type of dose-response
ENERGY FIELDS
relationship (Figure 29-2). Such a dose-response
Although MRI is considered safe, radiobiologists
will be busy for many years working to stretch

Response
Response

a
Dose
FIGURE 29-1  The dose-response relationship for ionizing Intensity
radiation is linear. It is also nonthreshold, suggesting that It
even the smallest radiation dose carries a risk, albeit insig- FIGURE 29-2  The response after exposure to the energy
nificant. The value a represents the natural incidence of fields of magnetic resonance imaging is threshold and
the response. nonlinear in form.
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 407

UNSAFE

Intensity

SAFE

IT

MRI

Time
FIGURE 29-3  Individually, the magnetic resonance imaging energy fields follow this type of time-intensity relationship
to produce a human response.

their ability to detect human responses to the

TABLE 29-1 Principal Mechanisms of
energy fields used. It is certain that the delayed
Interaction of the Three
determination that x-ray exposure can cause Magnetic Resonance
cancer and leukemia will not be repeated with Imaging Energy Fields
MRI, because the energy fields are nonionizing. with Tissue
The potential health hazards of MRI lie in the
static magnetic field (B0), the transient gradient Magnetic Resonance Mechanism of
magnetic fields (BSS, BΦ, BR), the radiofrequency Imaging Energy Field Interaction
(RF)-pulsed electromagnetic field, or a combina- Static magnetic field (B0) Polarization
tion of these. Radiobiological data already show Transient gradient magnetic fields Induced currents
that, individually, such fields are innocuous at the (BSS, BΦ, BR)
levels currently used for MRI. RF field (RFt) Thermal heating
In addition to following a threshold intensity-
response relationship, these fields exhibit a
time-intensity relationship (Figure 29-3). The IT
Static Magnetic Fields
applies to continuous exposure. Above this
threshold, shorter exposure times produce the The static magnetic field is probably the best-
same response. What is not fully understood is recognized MRI energy field. Although people
whether these MRI energy fields have adverse cannot sense a magnetic field, they know that it
health effects when combined. exists because of forces that attract and repel
Before there is a discussion of suspected bio- small magnets, such as kitchen latches and other
logical effects of MRI in human beings, an exam- everyday permanent magnets. Consequently, it is
ination of each of the MRI fields and their also clear that one interaction between a static
potential effects on other biological systems is magnetic field and matter results in a force that
necessary. The three MRI energy fields interact can turn certain types of matter (ferromagnetic
with matter differently (Table 29-1). material) into projectiles or missiles.
408 PART VII  Safety

Additionally, tissue can be rendered weakly

magnetic, or polarized, by exposure to a strong
static magnetic, field (B0). This induced tissue
magnetism cannot be sensed. Furthermore, in the
presence of a static magnetic field, tissue magne-
tization (M) can be measured only with great
difficulty in MRI. Once removed from the influ-
ence of a static magnetic field, tissue relaxes to
its normal state without a hint that it was previ-
ously magnetized.
There is no scientific evidence to show that
intense static magnetic fields produce harmful
effects in mammals. Mechanisms have been sug-
gested and subjected to considerable investiga-
tion but the result remains negative.
There are few studies of the effects of
long-term exposure of mammals to intense
static magnetic fields that use proper controls FIGURE 29-4  It has been shown that magnetic deposits
in the head of the pigeon allow it to know which direction
and large sample sizes. Some experiments have
is north.
been designed to detect growth abnormalities,
biochemical disruption, and malignant disease
induction in rodents at magnetic field strengths Monkeys who were exposed to a static magnetic
up to 5 T. All such observations have been field of 7 T for up to 1 hour experienced a tran-
negative. sient decrease in heart rate and sinus arrhythmia;
Other reports deal with the effects of intense a later study failed to demonstrate these effects
static magnetic fields on cultured cells. These with a limited exposure of 15 minutes at 10 T.
reports do not suggest any adverse effects on cell The evidence for static magnetic field effects
growth for magnetic fields up to 7 T. Such exper- on the developing embryos of pregnant mice
iments involve conventional culturing of human who were exposed at various times during gesta-
fibroblasts. Lengthening the exposure time does tion did not consistently show alterations in
not produce a positive response either. the offspring when compared with nonexposed
The navigational ability of adult bees and control mice. These studies used acceptably large
pigeons is considered to be mediated by granules sample sizes and were repeated with similar neg-
of magnetite deposits in the abdominal region ative results.
and in the head, respectively (Figure 29-4). These
magnetic deposits serve as a built-in compass, yet
Transient Magnetic Fields
experiments have not been conducted to deter-
mine whether these characteristics can be dis- The gradient magnetic fields used to spatially
turbed by intense static magnetic fields. Such encode the magnetic resonance (MR) signals vary
experiments are not pertinent anyway because with time. These are often termed moving mag-
human beings do not rely on an internal compass. netic fields or time-varying magnetic fields. The
Some speculation exists concerning bioef- preferred term is transient magnetic field.
fects of intense magnetic fields on growth rate, A transient magnetic field is able to induce an
mutation, fertility, and blood cell count in mam- electric current. For example, an electric current
mals. Growth rate and mutation effects have can easily be induced in a metal conductor. This
been reported in animals and yeast, but the is the basic principle of electromagnetic induc-
results have been equivocal or not reproducible. tion that generates and detects the MR signal.
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 409

Theoretically, transient magnetic fields can The electric field is generally well shielded
either stimulate or impair electric impulses along from the patient’s body by the placement of
neuronal pathways in tissue. The main influence guard rings in the RF coil design. The magnetic
of this is the time rate of change of the magnetic component of the RF field is required for imaging.
field expressed as dB/dt and measured in T/s. For However, in addition to flipping proton spins, it
example, a transient magnetic field of 3 T/s induces eddy currents within the conductive
results in an induced electric current density of tissues of the body. This results in heat, which
approximately 3 µA/cm2 in tissue. increases with increasing frequency.
A current density of 3 µA/cm2 is capable of A rise in body temperature has been reported
producing involuntary muscle contraction and with exposure to RF fields; the absorbed energy
cardiac fibrillation in experimental animals. causes this temperature increase. In turn, the
However, the time of application was far greater RF frequency, exposure time, and mass of
than that experienced in clinical MRI. the exposed object determine the amount of
absorbed energy.
Transient magnetic fields can induce neurologic The RF frequency influences energy absorp-
signals. tion because it is wavelength dependent. Energy
is most efficiently absorbed from an RF emission
Because some cells and tissues are electric con- when the tissue size is approximately one half
ductors, the transient gradient magnetic fields of the RF wavelength. The wavelengths associ-
may induce or interfere with normal conduction ated with 10 MHz and 200 MHz are 30 m
pathways. Depending on the intensity of the and 1.5 m, respectively. Shorter wavelength RF
induced electric current, the normal function of results in more superficial heating and less
nerve cells and muscle fibers may be affected. The penetration.
threshold current density of transient magnetic Commercial microwave ovens operate at
fields depends on the tissue conductivity and the 2.45 GHz. What is the wavelength of such radia-
duration of time the field is energized. tion? … 12 cm. What is the size of a hot dog? …
The maximum transient magnetic field does 12 cm.
not occur in the center of the magnet. There the Depending on the imaged region, when an RF
gradient magnetic fields are near zero; they pulse sequence is applied to a patient the peak
are maximum at the periphery of the imaging power can range to 16 kW. The energy returned
volume. from the patient is very weak; it is measured in
microwatts (µW). Some energy is lost to coil inef-
ficiencies and cable losses; some is reflected back
Radiofrequency Fields to the RF power amplifier and a load attached
The radiofrequency field (RF) used in MRI exists to the back of the coil.
in the approximate range of 10 to 200 MHz. Some energy is transmitted through the patient
This energy field consists of an oscillating electric without interactions, and some energy interacts
field and a similar orthogonal oscillating mag- with the patient and generates heat, causing a
netic field. These fields interact simultaneously possible rise in body temperature. The amount
with matter. The interaction manifests differently of absorbed energy depends on the transmitted
according to the nature of matter. magnetic field frequency, the total power of the
The electric field component can induce RF pulse, and the structure and conductivity of
heating in matter and induce high-frequency the exposed object.
electric currents, whereas the magnetic field com-
ponent acts as a rapidly transient magnetic field.
The physiologic measure of intensity of RF energy
This is relevant to MRI because it has the poten-
is the specific absorption rate (SAR).
tial for heating both superficial and deep tissues.
410 PART VII  Safety

The SAR has units of W/kg and may be The SAR is highest with pulse sequences that
expressed as an average over the whole body or require many 180° RF pulses, such as fast spin
any specific tissue. Thus there is a whole-body echo (FSE).
and a local SAR.
The SAR may also be expressed over a long
time or a single pulse. Specific absorption rate is The basal metabolic rate for a resting adult
a measure of the energy absorbed per unit time is approximately 1 to 2 W/kg. Anatomical abnor-
per unit mass. malities in the offspring of pregnant animals
Specific absorption rate is to MRI as the Gray have been reported when the SAR exceeded
is to ionizing radiation. The mathematical for- 20 W/kg during whole-body exposure. If the rate
mulation of SAR is extensive and unnecessary for of energy absorption exceeds the rate of heat
this discussion. loss, the body temperature increases. When such
Biological effects of RF are associated with the temperature increases last for long periods,
SAR. In turn, the SAR is related to the RF power adverse biological effects result.
density because it varies in time (temporal) and Exposure to microwave radiation, which
space (spatial). Recommended exposure limits involves frequencies higher than those used in
are expressed as power density. They are set MRI, causes a similar rise in body temperature.
approximately 100 times lower than levels In mice, such exposure has been associated with
known to cause a biological response. decreased fertility and testicular degeneration.
In a guideline issued in 1998, the Food and Such reports have not been confirmed by repeti-
Drug Administration (FDA) recommended that tive experiments.
equipment manufacturers in the United States
use an operating scheme put forth by the Inter-
national Electrotechnical Commission (IEC).
Combined Magnetic Resonance
Under the standard (IEC 60601-2-33, July 1995),
Imaging Fields
three modes of operation are established that In large-scale or long-term studies in experimen-
relate to RF heating and dB/dt. These are the tal animals exposed to the combined fields of
normal operating mode, the first level-controlled MRI negative results have consistently been
operating mode, and the second level-controlled reported. However, not all such experiments
operating mode; these are not directly related to involved exposure to all three MRI fields, and
any specific values but mark thresholds where the number of studies is limited.
concerns may arise. Mutagenic or lethal effects in Escherichia coli
The normal operation mode has been used for were not observed when the bacterial cells were
years for conventional imaging under the old exposed to typical MRI fields for up to 5 hours.
FDA guideline. The first level indicates when When human peripheral blood lymphocytes
there may be a situation of “undue physiological grown in tissue culture were exposed to MRI, no
stress” and requires medical supervision. The significant differences to chromosomal lesions or
second level indicates operation at a level that sister chromatid exchanges were found between
“may produce significant risk” and requires the exposed and control groups. Similar cytoge-
medical supervision and institutional review netic studies have been conducted with several
board approval. mammalian cells (e.g., HeLa cells and Chinese
MRI systems calculate the SAR. MRI tech- hamster cells) with MRI exposure. Negative
nologists cannot continue to image if the recom- results were reported in all studies in terms of
mended exposure limits will be exceeded. Always chromosome aberrations, sister chromatid
enter accurate patient weight to stay within the exchanges, and DNA synthesis.
recommended limits for SAR, even if these levels The effect of MRI on the early development
are lower than levels known to cause a response. of an amphibian system also resulted in the
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 411

absence of harmful bioeffects. These studies were

HUMAN RESPONSES
conducted with frog spermatozoa, eggs fertilized
TO MAGNETIC
during second meiotic division, and embryos
RESONANCE IMAGING
during cleavage. All were subjected separately to
30 MHz continuous wave RF in a static magnetic Before the advent of MRI, a considerable amount
field of 0.7 mT for 20 minutes. These specimens of research effort was put into studying these
were compared at similar stages with specimens electric, magnetic, and electromagnetic energy
in unexposed groups. Researchers compared the fields. With the introduction and clinical use of
damage in genetic material, interference with MRI and the properly focused concern for its
meiotic cell division, and impaired development safe application, more studies have been under-
of the embryos; no significant differences were taken at the cell, animal, and human level.
observed. This suggests that MRI exposure at Work in this area began in the early 1960s
these intensities does not cause detectable, adverse when some farmers in upstate New York com-
effects in the amphibian system. plained that cows grazing in pastures through
which high-voltage transmission lines ran pro-
duced less milk than cows in pastures without
Low-Frequency
these transmission lines. These reports were fol-
Electromagnetic Fields
lowed by additional concerns about aberrant
Though not directly applicable to the high- animal behavior, allegedly resulting from the
frequency electric and magnetic fields of MRI, energy fields of these high-voltage transmission
considerable public attention has turned to lines. These early reports led to investigations
the electric and magnetic fields associated with regarding the possibility of such effects.
60 Hz electric power distribution and use. Con- One study at Oak Ridge National Laboratory
sequently, low-frequency electromagnetic fields involved the construction of an enclosure to
(EMFs) also receive considerable scientific house eight cows at a time in controlled electric
attention. and magnetic fields (Figure 29-6). This and
The approximate electric and magnetic field other studies showed that the farm animals were
intensity associated with conventional electric unaffected by both the electric field and the mag-
power is shown in Figure 29-5. Note the relative netic field imposed by electric power transmis-
intensity of the magnetic field of the earth. sion lines.
Study results of these energy fields are largely The energy fields produced in clinical MRI are
negative. less intense than those associated with proximity
The Oak Ridge Association of Universities to electric power devices. However, the frequency
Expert Panel on EMF has expressed the follow- of the MRI RF is much higher than the frequency
ing on the subject: of 60 Hz electric power.
We have never stated that a causal association Two of the three MRI energy fields—transient
between EMF [electromagnetic fields] and cancer magnetic fields and RF fields—have been shown
is impossible or inconceivable; we have indicated to induce responses in human beings. Such
that the evidence for such an association is empiri- responses only occurred at exceedingly high
cally weak and biologically implausible. We
have not proposed that research concerning the intensities. There are no known effects from the
health effects of EMF be discontinued; in fact, static magnetic field.
we have indicated areas of some scientific interest However, transient magnetic fields induce
that warrant consideration for future research. magnetic phosphenes, stimulate healing in
However, given the decreasing resources available bones, and cause cardiac fibrillation. RF expo-
for basic health and science research, we believe
that in a broader perspective there are currently sure heats tissues, induces blood dyscrasia, and
more serious health needs that should be given forms cataracts. Note that all of these responses
higher priority. are acute.
412 PART VII  Safety

10,000

1000

Electric
field 440 kV Transmission line
intensity 100
(Vm−1)

10

Distribution line
1
0.1 1.0 10 100 1000 10,000
Distance from source (m)

1000

100
Earth

Magnetic 440 kV Transmission line

10
field
intensity
(µT)
1.0

0.1

Distribution line
0.01
0.1 1.0 10 100 1000 10,000
Distance from source (m)
FIGURE 29-5  Electric and magnetic field intensities associated with electric power.

There are no known long-term effects of MRI poor vascularity and a lower oxygen supply.
fields. Therefore it is suspected that the lens is more
susceptible to biological damage by RF energy.

The extent of SAR tolerance in a subject pri-

marily depends on oxygen supply and vascular-
Effects of Static Magnetic Fields
ity. The brain, kidneys, and liver are all tissues Even at 4 T, magnetic forces and torques on
with high vascularity. The lens of the eye has tissues are small when compared with ordinary
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 413

FIGURE 29-6  Enclosure for housing cows in a controlled electric field and magnetic field environment.

gravitational and inertial forces. The highest Electric stimulation of the sensory receptors of
static magnetic field that human beings can com- the retina as a result of transient magnetic fields
fortably tolerate is probably much more intense causes this phenomenon. The threshold for the
than 4 T. induction of magnetic phosphenes in human
When patients and volunteers have been beings is approximately 3 T/s at low frequencies.
imaged with these 4 T magnet systems, some Magnetic phosphene induction in the RF region
responses have been reported. These include has a threshold of approximately 20 T/s.
metallic taste, twitching, some disequilibrium, Bone healing accelerates if a low-frequency
and phosphene induction. Presumably, the tran- coil is positioned over a fracture. Although a
sient magnetic fields caused these responses. precise mechanism of action is unknown, this
When the examinations ended, the responses method works and is an accepted adjunctive
stopped. Researchers conducting these 4 T therapy. Treatments that last for several minutes
studies do caution that patient motion should be are repeatedly given during a 2-week to 5-week
minimized in these high-field imaging systems. course. Apparently, gradient magnetic fields
stimulate bone healing.
Ventricular fibrillation is another potential
Effects of Transient Magnetic Fields hazard of electric currents induced by transient
Magnetic phosphenes are flashes of light that can magnetic fields. This happens when the current
sometimes be perceived with the eyes closed. density in cardiac tissue is above 0.5 mA/cm2 and
D’Arsonval first reported magnetic phosphene applied for longer than 3 seconds. The threshold
induction in 1896. The introduction of MRI for such an effect is thought to be 0.1 mA/cm2.
rekindled research in this area. Such disruption of the heartbeat may cause a
414 PART VII  Safety

drop in blood pressure; however, ventricular

TABLE 29-2 Clinical Magnetic
fibrillation ceases and the heartbeat returns to
Resonance Imaging
normal when the induced electric current is inter- Exposure Limits
rupted. However, this response has been observed Recommended by the
only in patients with pacemakers. United States Center for
Each of these responses that can be induced Devices and Radiologic
by transient magnetic fields has a threshold that Health and the British
is a complex function of frequency, waveform, National Radiological
pulsatile nature, and duration. Presumably, each Protection Board
response follows the time-intensity relationship
shown in Figure 29-3. Recommended Limits
United Great
Effects of Radiofrequency Fields States Britain

The hazard of exposure to RF radiation is associ- MRI Energy Field

Static magnetic field <3.0 T <2.5 T
ated with heating. Thermal effects on tissue are
Transient magnetic field <3.0 T/s <20.0 T/s
related to frequency and waveform. The hazard
RF Field
is one of cooking, as in a microwave oven. Such Whole body <0.4 W/kg <0.4 W/kg
effects have been observed in both experimental Any 1 g of tissue <2.0 W/kg <4.0 W/kg
animals and patients.
Heating of avascular structures, such as the
lens of the eye, produces cataracts. This has been
demonstrated experimentally in animals and has patients have been imaged by MRI. The most
been observed in human beings. Some evidence predictable results of exposure to the MRI
suggests that ship-bound sailors who worked fields include an increase in temperature, physio-
near radio antennae and were exposed to high- chemical changes, and induction of electric
intensity RF subsequently had cataracts. currents.
Exposure to intense RF has been shown to None of these changes or any such adverse,
result in nonspecific blood changes, principally acute effects have been reported in patients
lymphocytic depression. This was thought to undergoing clinical MRI evaluation. However,
be the case some years ago when U.S. embassy as many as 5% of all patients experience
personnel in Moscow were alleged to have claustrophobia; another 10% sleep during the
been exposed to RF by sources outside the examination!
embassy. In light of the rapid progress in the application
Large-scale, epidemiologic studies of human of MRI in medical diagnosis, dose limits of
beings, with respect to the RF fields of MRI, have each component field of a clinical MRI system
not been done. To show any such effects, dedi- have been recommended by the U.S. Center
cated and experienced researchers would need to for Devices and Radiologic Health and by the
conduct a well-controlled, long-term study using British National Radiological Protection Board
a sufficiently large population of individuals (Table 29-2).
exposed to MRI. A prospective study of patients The American National Standards Institute
exposed to clinical MRI is currently lacking. (ANSI) standard for whole-body exposure in
comparison with the limitations that were used
by the former Soviet Union is shown in Figure
RECOMMENDED GUIDELINES 29-7. These standards acknowledge that biologi-
The first studies conducted on MRI patients cal responses are related to the power density
date back to early 1981. Since then, millions of and the frequency.
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 415

1000

100 ANSI standard for whole-body exposure

10
5

Power density in mw/cm2

1

.1

Russia standard, general population

.01

.001

.0001

.00001
.5 3.0 5.0 30 300 915 2450 6000
1500 3000
Frequency in MHz
FIGURE 29-7  Recommended limits of radiofrequency power densities are related to frequency.

TA B L E 2 9 - 3 Specific Absorption Rate TABLE 29-4 Recommendations of the

Levels from Which International Radiation
Radiofrequency Exposure Protection Association
Limits Are Derived (IRPA) for Exposure
Limits to Radiofrequency
Whole-Body Tissue Exposure Radiation
Population Exposure (1 g)
Occupational 0.4 W/kg 0.04 W/kg Power Density Limit (W/m2)
General public 0.08 W/kg 0.8 W/kg Frequency
Range (MHz) Occupational General Public
0.1 to 1 100 20
>1 to 10 100/f 20/f
In 1983 the International Radiation Protec-
>10 to 400 10 2
tion Association (IRPA) recommended interim >400 to 2 K f/40 f/200
guidelines on limits of exposure to RF EMFs. >2 K to 300 K 50 10
These guidelines are established for occupation-
ally exposed persons and the general public.
These limits are based on the SAR values shown
GENERAL SAFETY
in Table 29-3; the resulting RF exposure limits
CONSIDERATIONS
are shown in Table 29-4.
The basal metabolic rate for human beings is Radiologists use their knowledge of the potential
approximately 1 W/kg at rest. During exertion, human bioresponses to the energy fields of MRI
the metabolic rate may increase to 15 W/kg. to ensure safety for their patients and personnel.
416 PART VII  Safety

FIGURE 29-8  Crazed student technologists performing a routine physics experiment on a professor with
claustrophobia.

The following safety precautions are not dis- or otherwise confined space. The professor
cussed in a particular order. shown in Figure 29-8 is clearly claustrophobic.
True claustrophobes would probably not
even enter the MRI suite. Patients with mild
Patient Evaluation
claustrophobia need special attention to help
When patients arrive for an MRI examination, them through the examination. This special
they should complete a patient information form attention begins with the referring physician who
to make sure no contraindications are present. should prescreen all patients and counsel those
However, completion of the form is not enough who might be claustrophobic.
because patients do not know what is and what The MRI facility should be forewarned of a
is not important enough to keep them from possible claustrophobic patient so that more
having an MRI examination. The receptionist or planning and attention to patient preparation is
MRI technologist should spend sufficient time possible. The examination should be clearly
with each patient to explain the nature of the explained to the patient, and every effort should
examination and to determine the patient’s con- be made to comfort the patient before and during
dition (e.g., pregnancy, pacemaker, prosthesis, the imaging procedure.
metallic implant, occupational history). It is helpful to have a member of the family
or friend remain in the magnet room during
Patient screening is essential to ensure patient imaging. Usually, if a family member or friend
safety during MRI. maintains contact by holding onto the patient’s
foot and talking, the examination can be com-
Claustrophobia.  Patient anxiety can cause dif- pleted without incident.
ficulty in successfully completing an MRI exami- The development of open MRI has eased
nation. Claustrophobia accounts for most patient imaging of claustrophobic patients. Previous
anxiety during an MRI examination. open MR imaging systems were based on perma-
Claustrophobia is an excessive, sometimes nent magnet technology and had low field
morbid, fear of being enclosed in a tight, narrow, strength. Several vendors now offer open MRI
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 417

models based on superconducting magnet design should not proceed until determination that the
with B0 up to 1 T. clip is nonferrous.
Pacemaker.  Cardiac pacemakers are particu- Because many patients undergoing surgery
larly sensitive to magnetic fields. Pacemakers are will be MRI candidates in the future, surgical
designed to produce a small, electric voltage at protocol should require surgeons to positively
the cardiac frequency to maintain a proper identify all metal implants and pay specific atten-
heartbeat. tion to their ferromagnetism. The name of the
Integral to the pacemaker is a device called a manufacturer, model number, lot number, and
reed switch, which opens and closes in response serial number for any implanted device should
to a small electric signal. The reed switch can be be documented in the patient’s chart.
rendered inoperative, and the pacemaker fails in Implants identified as surgical steel, surgical
static magnetic fields as low as 2 mT. stainless, or stainless steel are not necessarily
nonferromagnetic. Many types of stainless steel
Patients with pacemakers must be excluded from are ferromagnetic.
the MRI suite and must not be examined by MRI. The magnetic properties of all such devices
should be determined directly and not assumed.
Because of the sensitivity of pacemakers to Even published lists by manufacturers of nonfer-
static magnetic fields, each MRI facility should romagnetic clips may be inaccurate. It may be
maintain a controlled environment outside of appropriate to delay an MRI examination until
a 0.5 mT fringe magnetic field. Controlling the precise magnetic state of any implant is
such an environment may require additional determined.
magnetic shielding or safety exclusion fences. No Although the warning is not particularly spe-
public passageway or other public access should cific, the FDA requires MRI systems to be labeled,
be possible in a fringe magnetic field exceeding “This device is contraindicated for patients who
0.5 mT. have electrically, magnetically, or mechanically
Biomedical Implants.  There have been several activated implants. For example, cardiac pace-
tragic accidents and one death because an makers because the magnetic and EMFs pro-
implanted surgical clip came under the influence duced by the MR device may interfere with the
of the intense static magnetic field of an MRI operation of these devices.”
system. The death involved an intracranial aneu- In addition to cardiac pacemakers, other items
rysm clip that twisted during an MRI and caused represent clear concerns and may contraindicate
the patient’s middle cerebral artery to bleed. This MRI (Box 29-1).
created a massive cerebral hemorrhage. Early in the use of MRI, prosthetic heart
If a surgical clip is ferromagnetic, it will expe- valves were considered a contraindication for
rience a torque induced by the static magnetic such examination. Valves implanted since 1982
field and twist in the tissue. The magnitude of are considered safe; earlier valves may also be
the torque is primarily determined by the strength safe, especially if imaged in a 0.5 T or less
of the external magnetic field; the degree of fer- imaging system. The forces created by the mag-
romagnetism; and the size, shape, and mass of netic field of an MRI system are small in
the clip. The potential for harm increases with comparison with those exerted by the heart.
an increase in any of these characteristics. When prosthetic valves are present, magnetic
Surgical clips that were made before 1980 uniformity is disturbed, and an image artifact
should be considered ferromagnetic. Since 1980, (usually absence of signal) occurs in the region
manufacturers of surgical clips have produced of the valve.
only nonferrous surgical clips. If it is determined In addition to the possibility of mechanical
during the patient interview process that a clip is displacement is the possibility of a biomedical
present in a critical tissue, then the examination implant dislodging. The possibility for the RF
418 PART VII  Safety

BO X 2 9 - 1 Devices That May Be TABLE 29-5 Typical Decibel Levels

a Contraindication for for Familiar Sounds
Magnetic Resonance Imaging
Decibel Level Description of Sound
Aneurysm clips 130 Threshold of pain
Biopsy needles 120 Front row seats at a Metallica
Bone growth stimulators concert
Bullets 110 Fourth-of-July fireworks
Cardiac pacemakers 100 Texas A & M University vs University
Cochlear implants of Texas football game
Dental implants 90 Hard Rock Cafe
Halo vests 80 Loud radio
Heart valve prosthesis 70 Motorcycle ride
Hemostat clips 60 Conversational speech
Implantable drug infusion pumps 50 Crackling fire
Internal defibrillators 40 Typical home
Intravascular stints 30 Bambi’s babies
Neurostimulators 20 Mirror Lake campsite
Occular implants 10 Final exam classroom
Orthopedic implants 0 Ants dancing
Penile implants
Vascular clamps

difficulties, and anaphylactoid reactions. Essen-

tially, those responses attributed to iodinated
field to induce electric currents or produce focal MRI contrast injections only occur at a lower
heating at the implant also exists. Neither of frequency.
these potential responses has been exhibited The use of MRI contrast agents with pregnant
either experimentally or clinically. patients requires special attention and should
Contrast Reactions.  At this time, four products probably be avoided unless the added image
are available to enhance the contrast of an quality is clearly justified. Apparently, the
MR image. Each of these is a chelate of gado- Gd-DTPA easily crosses the placenta, is swal-
linium, a paramagnetic element. Magnavist, the lowed by the fetus, and passes through the fetal
Berlex product, was the first on the market, and urinary tract. Research results in this area are
the competitors include ProHance, produced by still equivocal.
Bristol-Myers/Squibb; Omniscan, produced by Auditory Concerns.  When an electric current is
Sterling-Winthrop; and Dotarem, produced rapidly applied to a gradient coil, the coil heats
by Guerbet Laboratories. and expands. When the current is off, the coil
Each of these effectively enhances tissue con- contracts. This alternate expansion and contrac-
trast. Also, each has been scrutinized for safety. tion causes the banging noise heard during MRI
Iodinated contrast agents used in x-ray imaging examinations.
result in approximately one serious complication This banging noise created by the switching of
per 100,000 doses administered. The perfor- gradient magnetic fields can approach levels up
mance of these MRI contrast agents is somewhat to 95 dB. This is not far below the threshold of
better. pain, 120 dB, so ear protection may be required
Complications that may occur include nausea, (Table 29-5).
pariorthodal edema, nasal and ear lobe swelling, Some patients undergoing an MRI have expe-
neurential spasm, total body erythema, breathing rienced a temporary hearing loss. Hearing
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 419

protection is readily available with earplugs of

BOX 29-2 Objects Reported as
various designs. Such protective devices should
Projectiles in a Magnetic
be offered to all patients. Resonance Imaging Facility
There are a number of music headset systems
designed to muffle gradient noise and soothe Ankle weights
patients at the same time. Such antinoise systems Calculator
are highly recommended. Cigarette lighter
Clipboard
Floor buffer
Ferromagnetic Projectiles Forklift tines
Clearly, the most recognizable potential hazard Gurney
Hairpin
to an MRI examination is the potential for injury
Hearing aid
by projectiles. Regardless of size, ferromagnetic
ID badge
material can come under the influence of the Insulin pump
static magnetic field and become a projectile or Jewelry
missile. Key
Because of the attractive force of the primary Knife
magnetic field, ferromagnetic materials acceler- Mop
ate to the bore of the magnet if they are not Mop bucket
properly shielded. If an employee or a patient is Nail clipper
in the way, injury can occur. Many such injuries Oxygen tank
and two deaths have been reported. Box 29-2 Pager
Pole for intravenous bag
lists various items that have been reported as
Pulse oximeter
projectiles in MRI.
Scissors
It is essential for MRI administration to Steel-tipped shoes
conduct safety in-service sessions for other hos- Stethoscope
pital personnel. Nurses, porters, and housekeep- Traction sandbag
ing personnel can most often become victims of Vacuum cleaner
the adverse effects of the static magnetic field. An Various tools
incident can occur when least expected because Watch
of inattention or lack of information. Wheelchair
Writing pen

Ferromagnetic projectiles are a significant MRI

hazard.
rolled up to the MRI couch, and the oxygen tank
Be on the alert for such objects as scalpels, is sucked into the magnet!
oxygen tanks, sandbags, and jewelry. Nurses and Regardless of the number of in-service pro-
physicians often carry surgical instruments; grams and the amount of caution taken to train
patients often wear jewelry. These small items personnel, floor-cleaning instruments (e.g.,
can be particularly hazardous because of the buffers and polishers) seem to have an affinity
potential for puncture wounds. for MRI magnets. Several cases have been
Patients are often taken to MRI for examina- reported where such appliances have been
tion on a gurney and covered by a sheet. There attracted to and trapped in the magnetic field.
have been several reported instances in which Sandbags used for positioning and stabiliza-
transporters placed oxygen tanks on the gurney. tion of patients often contain not only sand but
Imagine everyone’s shock when the patient is also iron pellets.
420 PART VII  Safety

ate if it is likely that the information gained will

PREGNANCY AND MAGNETIC
materially affect the management of that patient.
RESONANCE IMAGING
MRI may also be appropriate for examina-
There is no evidence to suggest that MRI is tions of the brain and spine during pregnancy
harmful to a pregnant patient or a fetus. Never- when other modalities involve higher risk. Even
theless, the use of MRI during pregnancy is x-ray examination may be appropriate for preg-
approached with great caution. Researchers con- nant patients when considering the known risk.
tinue to study potential bioeffects. Most patients who receive radiation therapy
One potential problem associated with MRI treatments to the trunk of the body and subject
during pregnancy is heat. Although it is highly the fetus to considerable levels of ionizing radia-
unlikely that a fetus would sustain a measurable tion do carry the fetus to term.
elevation in temperature, it is known that heat Pregnant technologists operating MRI systems
is teratogenic. However, this property is no dif- are perfectly safe when working in the environ-
ferent from that created by exposing the preg- ment of the MRI. Such technologists are exposed
nant woman to a steam room, a hot tub, or a only to fringe fields and not to the principal
heating pad. imaging fields. The RF and transient magnetic
fields are essentially confined to the imaging
MRI-induced heat is a potentially harmful agent. volume of the patient and actually exist only
during imaging.

If MRI is the only imaging modality to provide Occupational exposure to the energy fields of
necessary information to the clinician for the MRI is safe.
proper managing of the pregnant patient, then
the examination should be conducted because The static magnetic field outside the imaging
clinical techniques do not significantly elevate volume is measured in millitesla, which is well
tissue temperature. If the attending physician below the intensity suspected of being capable of
needs the information from the MRI examina- producing an effect. MRI technologists are
tion, the examination should proceed after the exposed to a static magnetic field several orders
physician consults with the radiologist to select of magnitude lower than that to which a preg-
the appropriate pulse sequence. nant patient might be exposed. Consequently,
If an ultrasound examination can provide the there are no known or suspected hazards of oper-
required information, that should be the exami- ating an MRI system; therefore, restrictions on
nation of choice. On the other hand, if an exami- pregnant MRI technologists or radiologists are
nation is essential and the choice is between unnecessary.
x-ray examination and MRI, then on the basis An ongoing mail survey of MRI technologists
of safety alone, MRI should get the nod as the conducted by Kanal and Shellock has provided
appropriate examination. support for the presumed safety of such opera-
As with diagnostic ultrasound examination, tors, even during pregnancy. The miscarriage
the stage of pregnancy should not affect the deci- rate, frequency of complication, and condition of
sion to be examined with MRI. If an ultrasound the newborn were the same for MRI technolo-
examination of the abdomen or pelvis of a preg- gists as for the general population. There is no
nant patient has been unsuccessful in identifying measurable effect on any newborn that is attrib-
suspected abnormalities, MRI may be appropri- utable to the mother working with MRI.
 CHAPTER 29  Biological Effects of Magnetic Resonance Imaging 421

6. List the general concerns and describe the

CHALLENGE QUESTIONS
nature of each for the safety of the MRI
1. What type of dose-response relationship patient.
best applies to the energy fields used in 7. Are there any potential health effects from
MRI? the gradient magnetic fields?
2. Discuss the special concerns we have regard- 8. Define SAR and identify the recommended
ing pregnant MRI technologists. maximum whole-body exposure level.
3. Tissue responses to the energy fields used in 9. What is the principal potential response to
MRI are said to be deterministic. What does patient irradiation with RF energy?
that mean? 10. What units are used to measure the three
4. What is the one potential health hazard for energy fields involved with MRI?
pregnant MRI technologists?
5. What is the possible consequence of pro-
longed exposure to an intense B0 field?
 CHAPTER

30 
Managing a Magnetic
Resonance Imaging System

OBJECTIVES
At the completion of this chapter, the student • Identify the practical time per patient
should be able to do the following: schedule.
• Identify categories of equipment not permitted • Describe the necessary contents of a patient
in the magnetic resonance imaging (MRI) information request form.
examination room. • Describe those patients who should be
• Describe the start-up and continuing staffing examined by MRI.
requirements of an MRI facility.

OUTLINE
Ancillary Equipment Cryogen Replacement Zoning
Human Resources Magnet Quench Magnetic Resonance
Staffing Quality Control Personnel
Scheduling Safety Patient Preparation
Imaging Protocols Visitors’ Effect on Magnet Occupational Groups
Imaging System Maintenance Magnet Effects on Visitors Surgical Clips and
Preventive Maintenance Magnetic Resonance Safety Prostheses
Repairs Policies and Procedures Pregnant Patients

KEY TERMS
Accreditation Cryogen Preventive maintenance
Certification Informed consent Site selection
Continuing education Nonmagnetic

The time to begin preparing for the daily opera- as all that needs to be accomplished before the
tion of a magnetic resonance imaging (MRI) first patient enters the magnet is considered. A
facility is at least 1 year before the scheduled fair amount of research and thought are required
turnover date from the manufacturer. The reasons in producing a workable clinical environment
for this long planning time become clear as soon (Figure 30-1).

422
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 423

Acceptance testing
Inservice seminars
Staff training
Install Imager
Identify professional staff
Hire technical staff

Order ancillary equipment

Prepare site
Hire site manager
Design facility

Identify site
Select imager

0 3 6 9 12 15 18 21 24
Time (months)
FIGURE 30-1  The steps and time sequence for establishing magnetic resonance imaging (MRI) have been compared
with climbing a mountain, though MRI may take more time.

For the establishment of a properly function-

BOX 30-1 Ancillary Equipment
ing MRI facility, the time from conception to
Necessary for Magnetic
operation may take up to 2 years. In Chapter 13, Resonance Imaging Start-up
it is pointed out that the first step in this process
is selection of the imaging system. Then, site Fire extinguisher(s)
selection and design of the facility are the next Gurney(s)
processes. Approximately 1 year before opera- Wheelchairs(s)
tion, attention should be focused on personnel, Intravenous pole(s)
equipment, and administration. Step stool
First, all of the necessary ancillary equipment Oxygen tank (if there is no piped-in oxygen)
must be identified and purchase orders submit- Wooden chair(s)
ted. Then, a competent staff must be hired and Tools
Respirator (that will work in the magnetic field)
sufficient time allowed for adequate training
Stethoscope
of these new employees. Staff scheduling must Plastic bucket, mop
be arranged and imaging protocols established; Central vacuum (if there is carpet)
service and safety arrangements are also Patient monitoring (e.g., blood pressure, pulse rate)
necessary.

anesthesia, ancillary equipment becomes more

ANCILLARY EQUIPMENT sophisticated.
Routine daily magnetic resonance (MR) opera- Considerable thought should be given to ser-
tions require a certain level of ancillary equip- vices that will be offered at a particular site, and
ment support. As patient care becomes more then appropriate selection of equipment can be
and more intense, such as with sedation or made. Box 30-1 lists the equipment necessary for
424 PART VII  Safety

TA B L E 3 0 - 1 Representative Staffing for Start-up of a Magnetic Resonance

Imaging Facility

Number
Staff Member Hospital Clinic Duties
Radiologist 2 2 Image interpretation
Imaging technologist 4 2 Magnetic resonance imager operation
Receptionist 3 2 Scheduling, filing, typing
Radiology nurse 1 0 Patient support

the initial start-up and operation of most MRI fire extinguishers should be kept in the area to
systems. This list is offered as a planning guide. prevent any unfortunate accidents during an
emergency; a local supply company should be
Everything that goes into the magnet room must consulted for this.
be nonmagnetic. Apparatus is required for monitoring sedated
patients, elderly patients, and any patient with
Nonmagnetic equipment for use with an MRI impaired communication skills. One sedated
system is now readily available from several patient each day is a representative average.
suppliers. The primary materials used in such Minimal patient monitoring requires nonmag-
equipment are sturdy plastics, aluminum, wood, netic devices for assessing blood pressure, pulse
and high-grade stainless steel. Other metals and rate, and blood oxygen level.
alloys are also nonmagnetic, but it is always When considering the purchase of ancillary
best to check any large metal object with a small equipment, the administrator should not over-
magnet before it enters the imaging room. look the role played by small equipment in
Nonmagnetic gurneys are available from the front office area of a busy MRI facility.
many x-ray supply companies. Nonmagnetic Today’s competitive world requires an electronic
stretchers are more difficult to find and consider- office for general communications and for prompt
ably more expensive but are more versatile and magnetic resonance (MR) report delivery and
adaptable because they have more options, such filing.
as a raised head and adjustable height. Nonmag-
netic wheelchairs are also available, as are non-
HUMAN RESOURCES
magnetic intravenous poles and step stools.
Stainless steel or aluminum oxygen cylinders,
holders, and respirators are required. The
Staffing
cylinders can be obtained from either a local Staffing differs among facilities and also between
gas supply company or the cryogen supplier. a hospital and an imaging center. The number of
Specify that the tanks be labeled “nonmagnetic” staff depends principally on the patient load.
and verify that replacement tanks are the same. Table 30-1 identifies a representative start-up
A respirator that works in the magnetic field staff for both a hospital and an imaging center.
is primarily plastic with no metal gauges. After a few months of operation, it may be
The fittings should be nonmagnetic brass or necessary to increase hours and days of imaging.
aluminum. Many MRI facilities eventually operate 18 hours
Tools can be purchased from the MRI vendor a day, 6 days a week. This creates a great demand
or other companies that provide nonmagnetic for MRI technologists and radiologists to perform
supplies for research laboratories. Nonmagnetic complete diagnostic examinations in the most
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 425

efficient and timely manner. It may be necessary Strong signal intensity results in a bright
to schedule staggered shifts rather quickly after image, and weak signal intensity results in a dark
start-up. image. A complicating factor is that MR signal
What training and experience should be intensity can be completely reversed depending
required for MRI operation? Increasingly, more on the nature of the MRI pulse sequence.
formal programs are available to train competent Becoming comfortable with and understand-
technical personnel, specifically in MRI. In ing the physical principles of MRI are another
addition, many continuing education seminars matter. The technical staff must develop the nec-
are offered. Training provided by MRI vendors essary competence to exercise independent judg-
is usually excellent but tailored to specific ment about imaging techniques as they operate
equipment. the MRI system. Each staff member must receive
continuing education on a regular basis, with the
support of a site administrator who remains
Operation of an MRI system after only on-the-job
committed to continuing education.
training is unacceptable.

Scheduling
Radiologic technologists with extensive
experience in computed tomography (CT) are Experience has shown that approximately 30 to
an obvious choice to be MRI technologists. 45 minutes per examination should be allowed
Such individuals have already demonstrated initially. Early experiments with radiofrequency
competence as imaging technologists by complet- (RF) pulse sequences and patient positioning
ing a minimum 2-year training program and require considerable time. However, with experi-
passing the examination of the American Regis- ence the average examination time can be short-
try of Radiologic Technologists (ARRT). Their ened considerably. At that time, the scheduling
experiences in patient care, patient positioning, of technologists can be somewhat altered. Fast
anatomy, and computer operation can shorten imaging pulse sequences that shorten examina-
the learning time required for operating an MRI tion times considerably are now routine.
system. Ultrasound and nuclear medicine tech- The extremely competitive market for MRI
nologists are also likely choices. mandates the ability to offer same-day service
For the technical staff to develop the necessary and expeditious report turnaround time. It is not
competence to not only operate the MRI system unusual for a patient to leave a site, having had
but also exercise independent judgment about the MRI examination completed, with a copy of
imaging techniques, they should receive appro- the films and at least a preliminary report. The
priate continuing education. It is very important radiologist must be in a position of readiness to
for MRI technologists to have a basic under- accommodate the referring physician by being
standing of what occurs during MRI so that they available for consultation.
can provide adequate physician assistance. Tech- Patient scheduling should not be so tight that
nologist certification in MRI by the ARRT is now those who arrive on time must wait. In today’s
required in keeping with appropriate standards competitive atmosphere, patients who arrive on
of care. time and are properly prepared should not have
The physical basis for MRI is totally different to wait.
from that for x-ray imaging. Thought processes
equating more dense tissue to white and less Any hint of quality medical care can be dashed
dense tissue to black on a radiograph cannot be by causing patients to wait.
applied to MRI. Thus MRI is much more
complex. Technologists and physicians need to Technologist Scheduling.  During the initial
think in terms of signal intensity. start-up phase, every aspect of the procedure is
426 PART VII  Safety

slow because of such factors as unfamiliarity enter the examination room as the previous
with operating controls, difficulty in setting up patient exits.
imaging protocols, and selection of appropriate Request Form.  No patient should be examined
hard copies. Depending on the institution, this without a proper request completed by the refer-
period lasts 1 to 6 months. The technologists ring physician (Figure 30-2). As much medical
should become involved and be a part of the final information as possible should be provided with
product. The entire department benefits from this the request. In addition to giving standard iden-
interaction. Time and effort expended initially tification data, the requesting physician should
will bring rewards later. provide pertinent information from the clinical
Patient Scheduling.  Seldom should it be history of the patient and specific information
necessary to repeat an examination of a patient. about the reason for MRI.
This not only increases the workload but also In particular, the referring physician should be
disrupts the schedule. Nevertheless, reexamina- required to determine and state that the patient
tion is sometimes required because of an inability has no intracranial aneurysm or bypass clips,
to complete the examination. cardiac pacemaker, artificial limb, metal prosthe-
Occasionally, upon review of the images sis, or metal fragments. If the presence of any
by the radiologist, the appearance of a question- such items is questionable, radiography or CT
able area requires reexamination. The patient is images should be obtained to confirm their
contacted, and additional imaging is carried out absence. This should be done before the MRI
as soon as possible. In general, 15 to 30 minutes procedure is scheduled.
of image time can be expected. The patient Consent Form.  When any hardware or soft-
should be handled carefully and should be made ware of the MRI system is not approved by the
to feel as comfortable about the return visit as Food and Drug Administration (FDA), the FDA
possible. requires that a consent form be obtained from
The magnet should not remain idle because each patient. Regardless, it is advisable to obtain
of misinformation or scheduling inefficiency. a consent form from each patient. Figure 30-3
All outpatients should be called the day before presents such an informed consent form. In addi-
their scheduled examination for confirmation of tion to providing general information and iden-
appointment. At that time, staff should verify tification, the form must state that the patient
that patients know the time of their appointment understands the nature of the examination and
and should give instructions and directions to the ensure that all the patient’s questions have been
MRI facility. Time should be taken to answer answered.
questions or discuss fears patients may have. Of particular importance, the patient should
Such attention will help tremendously in ensur- be required to respond to the same questions
ing their arrival on time. posed to the clinician regarding clips, pacemak-
ers, prostheses, and other metal devices. A basic
explanation of MRI should be provided to the
Every effort should be made to prevent the
patient. A brief and clearly stated pamphlet
“empty table syndrome.”
would be sufficient. The patient should be made
aware that MRI is a benign examination and that
It is helpful to schedule similar examinations there are no adverse effects.
back to back. If possible, separate segments of Patient Preparation.  Before the examination,
each day should be scheduled for head, body, and the procedure should be explained and the patient
surface coil imaging. This eliminates the need for told what the examination time will be and that
switching equipment or coils from patient to it is important to remain still and relaxed.
patient and saves considerable time. An efficient The patient should be assured that the exami-
facility should strive to have the next patient nation does not hurt but that loud thumping
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 427

FIGURE 30-2  A representative magnetic resonance imaging request form.

428 PART VII  Safety

FIGURE 30-3  A typical consent form to be completed by the patient.

noises may be heard. Eating and drinking should are properly screened and prepared for the MRI
be avoided for 2 hours before the examination examination. Screening begins with pamphlets to
for patient comfort for a standard MRI. Imaging educate the referring physicians, their staff, and
the gallbladder and pancreatic ducts requires patients.
avoiding anything by mouth (n.p.o.) for 6 to Patient education continues throughout a
8 hours. The patient should be reminded that series of interviews with the receptionist and
this is a diagnostic examination and not a technologists. Sometimes the radiologists are
treatment. also involved in these interviews.
Image quality can be substantially improved The top priority of the MRI technologist
and image artifacts significantly reduced if patients should be to educate the patient about the nature
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 429

TA B L E 3 0 - 2 Radiofrequency Pulse Sequences That Emphasize the Principal Magnetic

Resonance Imaging Parameters

MRI Parameter RF Pulse Sequence TR (ms) TE (ms) TI (ms) Flip Angle

PD Spin echo 2000 30-50
T1 Spin echo 500 Up to 20
T1 Inversion recovery 2000 30-50 100-500
T2 Spin echo 2000 80-150
T2 Gradient echo 100 10 20°

of the MRI examination to obtain the greatest is the primary responsibility of the technologist
cooperation from the patient. This is necessary to learn how to manipulate the operating console
for minimizing voluntary motion. controls to optimize each examination.
The patient should realize that it is important
to remain as still as possible during the imaging Imaging protocols should be continuously
process. This means that the patient should be reviewed and updated.
relaxed and should breathe normally. Earplugs
that attenuate the noise but allow loudly stated The number of available MRI pulse sequences
instructions to be heard are often helpful. is enormous. In general, the technologist selects
To allay patients’ fears of being forgotten once the pulse sequence that provides the most infor-
inside the magnet, a technologist should inform mation in the least amount of time. Often,
patients that they are kept under observation in two pulse sequences are used: proton density–
several ways and that they may signal to the weighted (PDW) images for anatomy and
technologist at any time. Microphones and even T2-weighted (T2W) images for pathology.
a mere gesture of a moving foot or hand to alert MRI is valuable because of the enhanced con-
the technologist are helpful. A marginally claus- trast resolution resulting from differences in the
trophobic patient may be encouraged to com- basic MRI parameters of proton density (PD),
plete an examination if brought out of the magnet T1, and T2. Most pathologic processes cause
after signaling for a short break. changes in T1 and T2; therefore pulse sequences
are generally tailored to emphasize such changes.
Table 30-2 shows ranges of values for repetition
IMAGING PROTOCOLS time (TR), echo time (TE), and inversion time
Once the MRI system is operational and quali- (TI) for several pulse sequences designed to
fied imaging technologists are on staff, it will emphasize a given MRI parameter.
be evident that the choices presented for collec- During spin echo (SE) imaging, increases in T1
tion of images from a patient are almost endless. tend to reduce signal intensity, and increases in
Most imaging systems are fairly versatile and T2 tend to increase signal intensity. Therefore it
allow for wide variations in pulse sequences. is rarely sufficient to use a single SE for imaging.
The ultimate goal is to find optimal pulse The technologist usually obtains an early SE
sequences for each patient and for each suspected image and a late SE image to avoid having an
pathologic condition. The MRI technologist abnormality appear isointense with normal tissue.
must work closely with the radiologist. Because T1 and T2 values for the brain are
Each manufacturer has its own recommended long and those for cerebrospinal fluid (CSF)
protocol for a given circumstance; the radiologist are even longer when compared with those for
has specific needs in mind, too. Nevertheless, it other body tissues, pulse sequences for imaging
430 PART VII  Safety

TA B L E 3 0 - 3 Suggested Protocols for Start-up Imaging

Examination TR (ms) TE (ms)

PDW 1400-2000 30
T1W 500-600 5-20
T1W (90-500 ms TI) 1500-2000 30
T2W 2000-6000 80-150
T2*W (20°-30° flip angle) 500-1100 30
Brain Transverse 2000 40/80
Spine Transverse 1000 35/70
Coronal 800 40/80
Sagittal 800 40/80
Chest Coronal 1000 30/60
Transverse 1500 30/60
Abdomen Transverse 800 30/60
Coronal 500 25/50
Pelvis Coronal 800 25/50
Transverse 1000 30/60
Extremities Sagittal 700 40/80
Coronal 700 40/80

of the central nervous system (CNS) are usually cost-effective than foregoing patient imaging
not appropriate for body imaging. Table 30-3 and the revenues generated through normal
presents representative imaging protocols. With operations.
experience, these protocols can be varied to
accommodate suspected pathology.
Preventive Maintenance
Technologists and radiologists will want to try
different pulse sequences until one is found that The sophisticated support systems for the main
works best for their magnet and patient popula- magnet require the most attention during regu-
tion. Technologists must be open to change and larly scheduled preventive maintenance. Failure
learn from experiences. of any electronic components can seriously affect
the quality of the image and can disrupt imaging
entirely. The magnet must be constantly fine-
IMAGING SYSTEM MAINTENANCE tuned to keep the quality of the MR signal high
An MRI system is a highly complex instrument to produce images with maximum signal and
that requires regularly scheduled preventive minimum noise.
maintenance and unscheduled service for repairs Several support systems for the magnet may
as needed. include items that require regular preventive
The warranty period for parts and labor for maintenance as specified by the manufacturer.
an MRI system should be at least 1 year. Longer These systems may include a chilled water system,
service contracts are available but at consider- a halon system, oxygen alarms, compressed air
able expense. Warranty service is expected to and pumps, and air conditioning systems. A
occur during normal working hours, though mobile MRI is equipped with generators that will
after-hour arrangements can be made at a higher also require service.
hourly labor rate. Even with this added expense, Many of these items have components that are
this situation is often preferable and more not normal off-the-shelf items and may have a
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 431

long lead-time for replacement. Therefore it may drives that exist with any other computer-assisted
be necessary to have some components available technology.
for replacement at the site to preclude unneces- Because of the sensitivity of an MRI system
sary downtime resulting from equipment failure and the complex problems associated with repair
and parts delivery. and service, several items are necessary to monitor
Before operation of an MRI system, the person the service provided. The following items provide
responsible for service and preventive mainte- some controls for the cost of repairs.
nance must be identified. That individual should Length of Time Required for Repair.  Time
then be charged with scheduling such mainte- spent for repairs should be due only to the failure
nance and deciding what backup parts and of the equipment. Service personnel should not
systems should be available on site. add time to the service report for such things as
Regular preventive maintenance on the MRI filling out the service report. All times should be
system includes diagnostics on the software and logged in and out by the service personnel.
checks of the RF subsystem. Changes in RF fre- Repair Completed, Problem Unsolved.  Many
quency do occur. Each system normally has a 10 hours may be spent trying to isolate the
to 20 Hz drift each day. The gradient sensitivity event that causes the problem. It is not uncom-
and offset can drift, and readjustment of the free mon to put the imaging system out of service
induction decay (FID) may be necessary to main- several times to correct the same problem. There
tain image quality. should be an awareness of how long this is hap-
Cryogen levels should be monitored, and any pening and expert repair help should also be
differences in cryogen boil-off should be noted requested.
for possible problems. For safety procedure Training and Competence of Service Person-
checks, the emergency shutdown and quench nel.  Most often, more than one service engineer
buttons should be verified as operational. is required to correct a problem. If additional
Occasionally, the patient couch should be engineers are on site for training or experience
taken out to check the cowling and the inner and are not productive, there should be no charge
bore for small ferromagnetic objects. for the additional personnel.
Before the MRI system is purchased, an esti- Communication and Headquarters for Ex-
mate of maintenance costs after the warranty tended Service Problems.  Some persistent
period has ended is helpful. Negotiations for problems require that assistance from the factory
service should be done as the system is purchased be available for resolution.
so that a realistic picture of operating expendi- Response Time to Reported Service Prob-
tures is available in the planning stages. lems.  Time for repair of problems should not
include the travel time of service personnel to the
facility. It is wise to negotiate for a minimum
response time for a service engineer to respond
Repairs to requests for assistance. If service is required
The components that require the most attention over the weekends, this item should be addressed
are the electronics. Once the magnet is energized, during the purchase negotiations.
it does not require service except for the super- Availability of Parts Locally.  Parts for the
conducting magnet, which requires cryogen MRI system should be available locally but rarely
replacement. Unless a quench occurs, the magnet are. Repair budgets should be increased signifi-
should not require repair. cantly after the warranty period to include
Nevertheless, the electronic components of replacement components. Knowledge of the
the MRI system are the primary elements of the magnet and the components is mandatory for the
repair and service process. The system experi- MRI administrator or technologist to control
ences the normal problems with disk and tape repair costs.
432 PART VII  Safety

Usually, nitrogen must be replaced regularly

Cryogen Replacement
and some systems require helium replacement
The superconducting magnet requires cryogenic annually. These liquids vaporize during filling of
gases for cooling. The principle of the supercon- the cryostat and during transport. Cryogens
ducting magnet is to create an environment that cannot be stored for more than several days in
does not require a continuous electrical energy the transport dewars or else the dewars may be
source. The 15 or more miles of windings in the empty when the magnet is ready for filling.
core of a superconducting magnet must be cooled Helium is obtained commercially from depos-
to less than 20 K. This is accomplished through its of natural gas in Texas, Kansas, and New
the construction of the sophisticated cryostat (see Mexico. Helium is separated by a process of
Chapter 10). freezing out the less volatile components, leaving
As long as the windings of the main magnet the helium gas that must then be liquefied. All
remain below approximately 20 K, the magnetic U.S. production of helium is under the control of
field is sustained at a constant field strength. the U.S. Bureau of Mines.
Cryogen must be replaced on a regular schedule, Spontaneous boil-off of cryogens consumes
and this is a technical problem that must be approximately 1% to 2% of the helium and 5%
considered before operation. to 7% of the nitrogen per day on a stationary
Certain magnets are constructed so that only magnet. Cryogen consumption for a mobile
liquid helium (instead of helium and nitrogen) is magnet can be considerably higher, depending on
required. The boil-off from the helium can be whether the magnet is moved frequently or oper-
captured, condensed to liquid, and reused. With ated in a stationary location.
this design, cryogen replacement costs are dimin- If the magnet is moved frequently and must
ished. Also, cryogen fills occur less frequently; be ramped up and down for moving, an addi-
therefore, service costs are reduced. Monitoring tional 1% is used during the ramping. Critical
of cryogen levels must be continuous. Abrupt cryogen levels are reached at approximately 50%
loss of cryogens results in a quench that can be of the total volume; below this level a quench is
disastrous to the magnet. possible.
Liquid nitrogen is a fairly stable cryogen with
which to work because it vaporizes at 77 K and
Magnet Quench
has a high surface tension. Liquid helium is not
so stable; it vaporizes at 4.2 K. Helium is an important choice for a cryogen in
It is necessary to obtain a contract with a magnets because of its chemical inactivity and its
cryogen supply company or the system manu­ low density. However, this presents special prob-
facturer to arrange delivery of cryogens and lems if the liquid helium begins to warm or is
service. Certain manufacturers arrange cryogen exposed to room temperature.
deliveries in addition to monitoring levels, order- Charles’ law states that at constant pressure,
ing proper quantities, and scheduling fills and the volume of a given mass of gas is directly
related service. proportional to the absolute temperature. The
Cryogen replacement is a critical operation. It importance of this law is apparent if the magnet
must be done by knowledgeable, fully trained should quench.
personnel. Regardless of who does the replace- During a quench, the liquid gases vaporize,
ment, proper safety procedures must be followed. causing the main magnet windings to rise in tem-
Safety glasses must be worn for eye protection perature and become electrically resistive. This
and heavy gloves for hand protection. All con- results in more heat, which results in more liquid
nective tubing and parts must be precooled to boil-off. The magnet windings cannot sustain the
avoid the introduction of heat into either the high current in the resistive phase and can be
cryogen transport dewars or the magnet itself. severely damaged. If this occurs, the magnetic
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 433

field intensity drops rapidly to zero and the

TABLE 30-4 Image Characteristics to Be
imaging system ceases to function.
Evaluated with Magnetic
Resonance Imaging Quality
A quench occurs when the temperatures of the Assurance Test Objects
liquid gases—nitrogen and helium—exceed
their respective boiling points. Image
Characteristic Test Object Configuration
It is possible for 100 to 150 1 of helium and Noise Uniform liquid bath
nitrogen to be vaporized in less than 1 minute. Uniformity Measure of signal throughout
This produces approximately 4000 fl3 of gas at uniform liquid bath
room temperature and normal pressure, which Spatial resolution Hole pattern or bar pattern
can quickly displace all the oxygen in the room. Contrast resolution Hole pattern in various
The recommendation is that superconducting thicknesses of plastic
magnets be equipped with an oxygen monitor Linearity Step wedge or various
that will sound an alarm if the oxygen level in paramagnetic samples
the imaging room drops below 140 ppm. Normal Sensitivity profile Ramped wedge or rod
Slice continuity Ramp or helix
levels of oxygen in air are 150 ppm (20%) by
volume.
This assumes even greater importance if the
design of the magnet room does not allow the Phantom images can be used to identify and
technologists to view the cryogen port directly. evaluate artifacts that occur. Artifacts attributed
However, most superconducting magnets are to the inhomogeneity of a living system may be
vented through a cryogen vaporization duct to in error and should be cross-referenced to a
the outside of the facility to prevent an undue phantom image for identification.
safety hazard to personnel and patients. Medical physicists have developed several
Should a quench occur, the magnet quickly MRI QC test objects. The American Association
loses its magnetic field and becomes inoperable. of Physicists in Medicine (AAPM) has a large
The magnet must then be recooled and restarted, compendium of such test objects available on
which can take considerable time, assuming request. Table 30-4 lists the important image
the magnet has not been damaged from the characteristics to be evaluated by such test objects.
overheating. Keeping the SNR at a level that produces
excellent diagnostic images should be a closely
Quenching is an extremely expensive and hazard- monitored procedure. Even with magnetic and
ous event and must be avoided. RF shielding, small subcomponents of the support
systems can introduce interference that causes
image degradation.
The American College of Radiology (ACR)
QUALITY CONTROL has developed a very effective accreditation
The image must be evaluated daily to maintain process. The ACR QC manual describes the
image quality. This is best accomplished by using special responsibilities for radiologists, technolo-
the first half-hour of the day for quality control gists, and medical physicists. This manual also
(QC). A standard reference measurement on an details the use of the ACR accreditation phantom
MRI phantom should be done daily and the (Figure 30-4). All MRI facilities should be ACR
images and data saved for at least 1 month to accredited.
evaluate subtle changes in image quality and The industry-estimated number of clinical
signal-to-noise ratio (SNR). A reduction in the MRI systems in the United States at the end of
SNR may be the first indication of a problem. 2001 was approximately 5000. As of March
434 PART VII  Safety

FIGURE 30-4  The American College of Radiology accreditation phantom. (Reprinted with permission of the American
College of Radiology, Reston, Virginia. No other representation of this material is authorized without express, written
permission from the American College of Radiology.)

2002, there were 2408 accredited MRI facilities The radiologist’s responsibility is chiefly
accounting for a total of 2873 magnets. There oversight. The radiologist must ensure that the
were approximately another 500 facilities some- technologists and medical physicist are properly
where in the accreditation process. Each site qualified and maintain satisfactory continuing
must have a daily QC program with the ACR education. Every 3 months the radiologist must
phantom (Table 30-5). The demand for these conduct a review of the QC program and main-
phantoms is increasing. tain records of that review.
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 435

TA B L E 3 0 - 5 Specific Tests Performed by the Magnetic Resonance Imaging Technologist

and Their Recommended Frequency as Listed in the American College
of Radiology Quality Control Manual

Test Frequency
Center frequency measurement Daily
Table positioning check Daily
Setup and imaging interface Daily
Geometric accuracy Daily
High-contrast resolution Daily
Low-contrast resolution Daily
Artifact analysis Daily
Film quality control Weekly
Visual checklist Weekly

FIGURE 30-5  A series of images acquired with the American College of Radiology accreditation phantom to evaluate
magnetic resonance imaging system performance. (Courtesy Geoff Clarke, San Antonio, TX.)

SAFETY
The medical physicist is responsible for accep-
tance testing and an annual imaging system per- Access to the magnet room must be strictly con-
formance evaluation that includes many precise trolled because of potential adverse effects of
measurements. A sample of images used to make visitors on the magnet and the magnetic field on
these measurements is shown in Figure 30-5. visitors. Signs that caution entrance to the magnet
436 PART VII  Safety

the magnet. The departments that require the

most attention in this regard are building ser-
vices, public relations, and personnel.
In-service training sessions with the supervi-
sors of the building services department and the
employees performing the work are vital to the
safety of the imaging system. For example, one
hospital had a central vacuum installed in the
area because the magnet room was carpeted.
Within a week of in-service session, a housekeep-
ing employee decided the central vacuum was too
troublesome and entered the magnet room with
a conventional vacuum cleaner. Fortunately, only
a chip was knocked off the cover of the magnet
cowling, which is where the vacuum cleaner
struck when it was pulled out of the housekeep-
er’s hands.
The public relations department must be made
aware of the nature of the MRI system because
the members will be touring the facility with visi-
tors. Another reported incident, much more
benign in nature, occurred after a magnet was
first brought up to field. There was a group of
medical administrative dignitaries visiting from a
nearby city who were in the preliminary stages
of planning for MRI. Signs were posted promi-
FIGURE 30-6  Warning sign for the entrance of the mag- nently as to what should and should not enter
netic resonance imaging suite. the magnet room. One visitor was warned by an
employee that he should not wear his analog
watch into the magnetic field. The man expressed
room must be prominently displayed in many little concern because it was a gold watch. The
areas in addition to the entrance door. An employee responded that although gold is non-
example of such a warning sign is shown in magnetic, the gears and springs of the watch are
Figure 30-6. not made of gold.
Many people read and then ignore such signs;
therefore additional measures must be taken to No metal can be allowed in the MRI room, unless
control entrance to the MRI examination room. screening shows it to be safe.
This is particularly important with permanent
and superconducting magnets because their mag- As stated earlier, MRI requires a different
netic field cannot be interrupted or turned off. thought process. If care and constant monitoring
are not practiced, the potential for hazard is
always present. Box 30-2 lists personal belong-
Visitors’ Effect on Magnet ings that should be removed from visitors or
When the MRI system first becomes energized, patients before they enter the MRI exclusion
in-service training sessions for many departments area.
in the hospital or clinic must be conducted. There All portable metallic devices must be identified
will be much curiosity and many rumors about as nonferromagnetic before being brought into
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 437

BOX 30-2 Personal Effects That Should BOX 30-3 People Who Should Not Be
Be Removed from Visitors Allowed Within the 0.5 mT
and Patients Before Exclusion Area
They Enter the 0.5 mT
Exclusion Area Patients with pacemakers
Patients with intracranial aneurysm clips
Analog watches Persons subject to uncontrollable seizures
Tape recorders
Magnetized credit cards
Calculators
Jewelry superconducting MRI system should be consid-
Shoes with metal supports ered hazardous. Box 30-3 identifies some people
Wigs who should not be allowed access to the imaging
Hairpins, barrettes
room or within the 0.5 mT exclusion area. Any
Dentures (when the head or cervical spine is being
patient, employee, or family member entering the
imaged)
magnet room should be properly screened.

MAGNETIC RESONANCE SAFETY

POLICIES AND PROCEDURES
After a number of serious accidents, including
one widely reported childhood fatality, the ACR
published recommended MR Safety Policies and
Procedures. These recommendations are based
on long-established radiation safety protocols of
identifying potentially hazardous areas, naming
responsible personnel, limiting access, and con-
tinuously documenting and reviewing.

Zoning
The ACR recommends identifying four zones of
FIGURE 30-7  A handheld magnet of at least 100 mT
strength can be used to screen objects before taking them
increasing potential hazard. Figure 30-8 shows
into the magnetic resonance imaging room. (Courtesy the location of these zones for a typical MR suite.
Emanuel Kanal, Pittsburgh, PA.) Zone One includes all areas that are freely
accessible to the general public. This would
include the outside areas surrounding a free-
the imaging room of the MRI facility. A toy standing MRI facility or the corridor of the hos-
magnet is not adequate for such a screening task. pital leading to the MRI department.
A very strong, handheld magnet of at least 100 Zone Two is that area where patients and visi-
mT is required (Figure 30-7). tors are controlled and supervised by MR per-
sonnel. This would normally be the reception
area and the patient preparation area.
Magnet Effects on Visitors Zone Three is the area where there is potential
The magnetic fringe field may have potentially for injury from ferromagnetic objects and equip-
harmful effects on visitors, patients, and accom- ment. Access to this zone must be strictly con-
panying family members. The fringe field of a trolled by MR personnel.
438 PART VII  Safety

Zone Four
MRI

Zone Three
(control room)

Computer
room

Zone Two
(patient dressing
and holding room)

Zone One
(reception)

Entrance

FIGURE 30-8  Magnetic resonance zones of increasing potential hazard for a typical magnetic resonance imaging
facility.

Such control should provide for physical Any area within Zone Three where the fringe
restriction with a pass-key locking system to magnetic field equals or exceeds 0.5 mT should
ensure that only MR personnel and properly be clearly marked.
screened and supervised non-MR personnel can
enter. Zone Three is usually the control room
and immediate space, such as the computer Zone Four is the MRI room itself. This zone
room. should be clearly marked with warning signs and
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 439

a warning light at the entrance … “The Magnet will have had extensive training in MR safety
Is On.” and MR emergency procedures. Level Two MR
The entrance to Zone Four should be visible personnel have unrestricted access to Zones
to MR personnel at all times to absolutely ensure Three and Four. No work restrictions are neces-
no unauthorized entry. Even during an emer- sary for Level Two MR personnel who are
gency, such as cardiac or respiratory arrest, pregnant.
unauthorized and unscreened individuals must
not be allowed access. It is the responsibility of
PATIENT PREPARATION
the properly trained MR personnel to stabilize
the patient, provide basic life support, and Although MRI is generally considered harmless,
remove the patient to Zone Two or lower. certain patients should not be imaged because
of possible adverse reactions to the examination.
It is not always easy to identify those patients
Magnetic Resonance Personnel who should be excluded because the criteria for
An MR Medical Director and MR Safety Officer exclusion are not obvious even to those familiar
(MRSO) should be named for each MRI facility. with MRI. The responsibility rests with the MR
Often the two will be the same person. technologist to recognize the potentially unac-
The MRSO must see that formal MR Safety ceptable patient.
Policies and Procedures are developed, imple-
mented, and updated as necessary. Any MR Scheduling personnel must carefully screen
safety incident or “near miss” must be reported patients.
to the MRSO, who will maintain a register of
such incidents. The register can then be used to Patients must remove all metallic personal
refine the MR Safety Policies and Procedures. belongings and devices before entering Zone
Three. This is ensured if the patient is provided
The MRSO will identify two levels of MR person- with a gown having no metal fasteners for the
nel and four safety zones. MRI procedure.
The use of metal detectors for patient screen-
The ACR recommends that all persons enter- ing is not recommended. Patients who have or
ing an MR facility be identified as falling into may have internal ferromagnetic objects, such
one of three groups. There are two levels of MR as metal shavings or surgical clips, must be
personnel and a third group of non-MR person- screened further.
nel, including the patient. If positive written identification and documen-
Non-MR personnel are those who have tation of internal metal are not possible, plain
received no MR safety training, such as the film radiography should be conducted. The
patient, visitors, and other hospital personnel. MRSO should make the final decision for exami-
Such individuals are not allowed beyond Zone nation of any patient with a questionable screen.
Two unless screened and supervised by Level
Two MR personnel.
Occupational Groups
Level One MR personnel are those who have
completed minimal MR safety training. This Members of several occupational groups should
training will allow them to work safely within be viewed with suspicion before MRI examina-
Zone Three. This group includes receptionists tion because they may have foreign metallic
and patient attendees. No other hospital employ- objects in the body. Often the presence of such
ees should have unrestricted access to Zone Three. objects is unknown.
Level Two MR personnel include the MR Auto mechanics, machinists, and welders
technologists and MR radiologists. These people come into contact with metal shavings and
440 PART VII  Safety

filings. Most people who work in such a capacity prosthesis. For example, it is possible to image
take protective measures, such as wearing safety the lumbar spine of a patient with bilateral hip
glasses and other protective apparel. Therefore prostheses. However, it is usually not possible to
they should have no history of injury. obtain a complete diagnostic examination of the
Exclusion of patients with intraorbital metal pelvis.
is common. This practice prevents injury caused
by magnetic metal torquing in the magnetic field,
Pregnant Patients
which has been reported.
Patients who have metal fragments in their There are no data currently available to suggest
bodies may come to the MRI facility with no that any harmful effects occur to a pregnant
history of foreign body injury. Usually, these patient or fetus. However, it is generally believed
fragments are fixed in subcutaneous tissue, but that until more research data are available, preg-
the possibility of a hazard always exists. nant patients should be imaged only when alter-
native procedures are considered ineffective.
Surgical Clips and Prostheses Pregnancy is not a contraindication to MRI at any
Numerous neurosurgical clips and implants have stage of pregnancy.
been tested, and some are magnetic. It is impos-
sible to recognize the type on any radiograph. MR contrast agents should not be used with
Even if all neurosurgical clips were changed to pregnant patients without the consent of the MR
nonmagnetic materials today, it would be impos- Medical Director or MRSO. Some adverse effects
sible to know in the future whether a patient had are possible; therefore each pregnant patient
one of the old magnetic aneurysm clips or a new should be individually evaluated.
nonmagnetic one.
CHALLENGE QUESTIONS
The patient with neurosurgical clips is not imaged
1. What characteristic must be kept in mind
unless the technologist is absolutely certain that
when equipping an MRI facility?
the clips are nonmagnetic.
2. A patient has a history of neurosurgery.
What should be done?
Middle ear prostheses are not considered haz- 3. What are the minimum qualifications for
ardous to patients but may somewhat compro- operation of an MRI system?
mise the image quality. Many but not all middle 4. What is the basis for an exclusion area sur-
ear prostheses that have been tested are nonmag- rounding an MRI suite, and what is the
netic. Some minor image degradation in the area intensity level of exclusion?
of the implant or prosthesis may occur. 5. Discuss some management features designed
Surgical clips and sutures in the abdomen are to provide the patient with efficient service.
more of a problem. There is no contraindication 6. Why is regularly scheduled in-service train-
to the use of MRI in the patient with surgical ing necessary for hospital personnel?
clips in the abdomen. On occasion, significant 7. Discuss some helpful ways to ease the claus-
artifacts indicating magnetic metal clips appear. trophobic patient through an examination.
Trial and error is the only way to determine 8. Discuss the recommendation for quality
whether abdominal surgical clips will produce control for an MRI facility.
artifacts that are objectionable. 9. What steps should be planned in the event
Joint prostheses are primarily constructed of an unintended quench?
of stainless steel and do not cause a problem 10. Why is an informed consent required from
unless the area of interest is directly next to the a patient undergoing MRI?
 CHAPTER 30  Managing a Magnetic Resonance Imaging System 441

"I really would like to accommodate you but you're going

to have to find another way to see if you have a heart."
 APPENDIX

A
The Bloch Equations

Within a few weeks after Felix Bloch’s announce- convention, the Cartesian coordinate system is
ment in 1946 regarding the discovery of nuclear oriented with the Z-axis parallel to the direction
induction, now known as nuclear magnetic reso- of the applied field (B0); the X- and Y-axes
nance (NMR),* he published a set of equations are mutually perpendicular to each other and to
describing NMR. The most remarkable feature the Z-axis. The laws of thermodynamics set the
of this set of three coupled differential equations maximum nuclear magnetization at a value M0,
is Bloch’s clear explanation. In addition to pro- and its magnitude (MX2 + MY2 + MZ2)1/2 can never
viding a description of the NMR experiments, exceed M0.
the Bloch equations are applicable to other areas Bloch postulated two relaxation times, longi-
of physics. Scientists compare the power of these tudinal and transverse. He associated longitudi-
equations with that of the Maxwell equations, nal with MZ and transverse with MX and MY, and
which relate electric and magnetic fields to elec- then he assigned them the symbols T1 and T2,
tric charges and currents, as well as forming the respectively. T1 and T2 are time constants for a
basis for the theory of electromagnetic waves. first-order kinetics process, which can be incor-
Even though NMR scientists in physics, chem- porated into a set of coupled differential equa-
istry, and engineering substantially alter the tions as follows:
mathematical notation of the equations, Bloch’s
set of equations, which predicts the behavior of dM X M X
= (1a)
nuclear spin systems in a magnetic field, is nearly dt T2
infallible. Despite the complicated appearance of
dM Y M Y
the Bloch equations, an intense study session can = (1b)
provide insight into the behavior of spins in mag- dt T2
netic resonance imaging (MRI). Familiarity with dM Z (M Z − M 0 )
= (1c)
spin behavior is critical to understanding the sig- dt T1
nificance of magnetic resonance measurement
parameters. The first two equations state that the trans-
Nuclear magnetization is often described as a verse components of the nuclear magnetization
vector in three-dimensional (3-D) Cartesian decay with the time constant T2. The third equa-
space with components MX, MY, and MZ. By tion states that the longitudinal component
builds up to the value M0 with a time constant
T1.
The following are some interpretations and
*The Bloch announcement from Stanford University was
one of two. A simultaneous discovery was made by applications of equations 1a, 1b, and 1c. Con-
Edward Purcell at Harvard University. The two men sub- sider a nuclear spin system at equilibrium created
sequently shared the 1952 Nobel Prize in physics. by placing a sample in a field B0 for a time t,

442
 APPENDIX A  The Bloch Equations 443

z z
M0 MY = M0 1.0
x x
0.8

MY as Fraction M0
y y
0.6

0.4
FIGURE A-1 
0.2
which is long in comparison with T1 and T2.
The following definition applies:
0.0
dM X dM Y dM Z 0 1000 2000
= = = 0 (2) Time (ms)
dt dt dt
That is, the components of the nuclear magneti-
zation do not change with time when the nuclear 1.0
spin system is at equilibrium.
Suppose by the intervention of an external 0.8

MX as Fraction M0
force, or pertubation, M0 is transformed from
0.6
the Z-axis to the Y-axis (Figure A-1). The time
history predicted by equations 1a, 1b, and 1c
0.4
can be used to interpret the Bloch equations
(Figure A-2).
0.2
The graphs in Figure A-2 have been generated
by stepwise solution of equations 1a, 1b, and 1c 0.0
with MY = M0, T2 = 100 ms, and T1 = 500 ms. 0 1000 2000
MY decreases with time constant T2 and Time (ms)
approaches zero after several T2 periods. MX
was initially zero and remains at that value. MZ,
1.0
also initially zero, increases with time until it
essentially becomes M0 after several T1 periods.
0.8
MZ as Fraction M0

Equations 1a, 1b, and 1c are the mathematical

statement of the T1 and T2 relationships for the
0.6
rotating frame of reference at resonance.
Note that numerical integration is a com-
0.4
monly used digital computing technique. The
name is formidable, but the technique does not
0.2
need to be. For the illustrations in this appendix,
assume a time change; calculate the difference in 0.0
the value of MX, MY, and MZ that occurs in the 0 1000 2000
time interval; correct the MX, MY, or MZ values; Time (ms)
and repeat the operation. The NMR enthusiast FIGURE A-2 
can accomplish several such steps on a handheld
calculator and then estimate the curves.
Remember that T1 must always be equal to
or greater than T2. If this restriction is violated,
the magnitude of the M0 vector can easily exceed
444 APPENDIX A  The Bloch Equations

the thermodynamic M0 limit. There is no known turned off, the behavior shown in Figure A-2
substance with T2 greater than T1. results.
Two factors allow for elaboration on equa- Until now, discussion has centered around the
tions 1a, 1b, and 1c. First, no method for direct rotating frame at resonance. When the rotating
observation of MZ exists. Quantum mechanics frame is not at resonance, it requires another
guarantees this fundamental truth. Second, a modification of the equations. The terms to be
physical means of perturbing nuclear spin systems included are the frequency of the rotating frame
at equilibrium has not been discussed yet. Intro- (ω) and the resonance frequency (ω0), or γB0. A
ducing a radiofrequency (RF) field in the XY difference (ω − ω0) corresponding to 1000 Hz
plane solves both of these limitations. means the magnetization rotates 1000 times per
An RF field can be an applied to secondary second. Mathematically, rotating magnetization
magnetic field (B1) along the X-axis of our coor- in the XY plane is incorporated by decreasing
dinate system. B1 rotates the vector M0 around MX and increasing MY with respect to time, as
the X-axis at a rate γB1, where γ equals the gyro- shown in equations 4a, 4b, and 4c. A compara-
magnetic ratio of the nucleus under observation. ble form exists for rotation in the YZ plane. The
Incorporation of this effect changes equations Bloch equations in the form of equations 4a,
1a, 1b, and 1c into the following equations: 4b, and 4c, which incorporate both of these
features, are reliable descriptors of the magneti-
dM X M X zation vectors in a sample during an NMR
= (3a)
dt T2 observation.
dM Y M Y
= + γB1M Z (3b) dM X M
dt T2 = − X + (ω − ω 0 ) M Y (4a)
dt T2
dM Z (M Z − M 0 )
= − γB1M Z (3c) dM Y M
dt T1 = − Y − (ω − ω 0 ) M Y + γB1M Z (4b)
dt T2
Equations 3a, 3b, and 3c are simply a math- dM Z (M Z − M 0 ) (4c)
=− − γB1M Y
ematical statement that any vector in the YZ dt T1
plane is rotated by B1 to a new location. Figure
A-3 shows the effect of ωB1 on a nuclear spin The offset terms only appear in MX and MY.
system that is initially at equilibrium. Numerical integration leads to the time history
The RF field B1 creates a motion of M0 around shown in Figure A-4.
the X-axis. The trajectory of M0 can be inter- In this case, the detected signal is a damped
rupted at will by turning off the RF oscillator. In sine wave for MX and a damped cosine wave for
this case, equations 3a, 3b, and 3c simplify to MY. This is indeed the observed quadrature NMR
become equations 1a, 1b, and 1c, with the com- signal.
ponents MX, MY, and MZ determined by the time The parameters MX, MY, and MZ are all
the RF field was turned off. After the RF field is affected by an RF pulse. These lead to three plots

z
x RF
along
y
X axis
1
γB

FIGURE A-3 
 APPENDIX A  The Bloch Equations 445

1.0
analogous to those in Figure A-4. B0 magnetiza-
tion varies with time and is affected by repeated
0.8
excitation. The solution to these three equations
0.6
MX as Fraction M0

by computer and a plot of the results versus time

0.4
permit the NMR user to visualize the events
0.2
during an NMR observation either spectroscopi-
−0.0
cally or by imaging.
−0.2
What faults exist in this analysis of NMR?
−0.4
One drawback is the assumption of a perfectly
−0.6
homogeneous magnetic field. In an inhomoge-
−0.8
neous magnetic field, each nucleus has a unique
−1.0
0 200 400 600 800 1000 offset frequency (ω − ω0). The NMR signal
Time (ms) behavior observed for this sample is the summed
behavior of the various nuclei. This process,
called forming the ensemble average, provides
1 the intellectual bridge between the behavior of
an isolated spin system and multiple spin systems
found in imaging applications. The ensemble
averaging procedure is not shown here, but it
MY as Fraction M0

can and does provide a cogent explanation for

0 the spin echo (SE) and image formation by mag-
netic field gradient modulation. A second omis-
sion is that the Bloch equations do not make a
provision for the spin-spin coupling interaction
that is always present in NMR spectroscopy.
−1 In conclusion, the versatile Bloch equations
0 200 400 600 800 1000 of nuclear induction, which describe how spin
Time (ms) systems behave in a magnetic field, lend them-
selves to modification and approximations
suitable for interpreting magnetic resonance
1.0 observations. A basic understanding of these
equations is essential before the novice can
0.8 become an expert in MRI interpretation.
MZ as Fraction M0

0.6

0.4

0.2

0.0
0 200 400 600 800 1000
Time (ms)
FIGURE A-4 
 APPENDIX

B
Additional Resources

Advance Newsmagazines American Society of Radiologic Technologists

www.advanceweb.com (ASRT)
www.asrt.org
Agfa HealthCare
www.agfamedical.com Applied Radiology
www.appliedradiology.com
AILab Co (Advanced Imaging Laboratory)
www.ail.co.kr Armed Forces Institute of Pathology (AFIP)
www.afip.org
Air Products and Chemicals, Inc.
www.airproducts.com/gases/keepcold.ht Association of Educators in Radiological
Sciences, Inc. (AERS)
American Association of Physicists in Medicine www.aers.org
(AAPM)
www.aapm.org Aurora Imaging Technology, Inc.
www.auroramri.com
American College of Medical Physicists
www.acmp.org Avotec, Inc.
www.avotec.org
American College of Radiology (ACR)
www.acr.org Berlex Laboratories
www.berleximaging.com
American Healthcare Radiology Administrators
(AHRA) Bracco Diagnostics, Inc.
www.ahraonline.org www.bdi.bracco.com

American Registry of Radiologic Technologists Braden Shielding Systems

(ARRT) www.bradenshielding.com
www.arrt.org
Cybermed, Inc.
American Roentgen Ray Society (ARRS) www.cybermed.co.kr
www.arrs.org
Diagnostic Imaging Magazine
American Society of Neuroradiology (ASNR) www.diagnosticimaging.com
www.asnr.org

446
 APPENDIX B  Additional Resources 447

Eastman Kodak Co. Lippincott, Williams and Wilkins

www.kodak.com/go/health www.lww.com

Elsevier Science Magnacoustics, Inc.

www.elsevier.com www.magnacoustics.com

ETS-Lindgren Medical Technology Management Institute

www.lindgrenrf.com www.mtmi.net

FONAR Corp. Medrad, Inc.

www.fonar.com www.medrad.com

Fujifilm Medical Systems USA, Inc. MRI Devices Corp.

www.fujimed.com www.mridevices.com

Gammex RMI MRI Safety, Bioeffects and Patient Management

www.gammex.com www.mrisafety.com

GE Medical Systems National Council on Radiation Protection and

www.gemedicalsystems.com Measurements
www.ncrp.com
Health Physics Society
www.hps.org Nuclear Associates/Inovision Radiation
Measurements/Syncor
Herley Medical Products www.victoreen.com
www.amtinc.com
Nycomed Amersham
Hitachi Medical Corp. www.us.nai.com
www.hitachimed.com
The Phantom Laboratory
IGC-Medical Advances, Inc. www.phantomlab.com
www.medadv.com
Philips Medical Systems
Image Systems Corp. www.medical.philips.com
www.imagesystemscorp.com
ProScan MRI Education Foundation, Inc.
Indiana University Radiology Education and www.proscan.com
Research Institute
www.indyrad.iupui.edu Radiological Society of North America
www.rsna.org
John Wiley and Sons, Inc.
www.wiley.com Resonance Technology, Inc.
www.mrivideo.com
Konica Medical Imaging, Inc.
www.konicamedical.com RT Image c/o Valley Forge Press
www.rt-image.com
448 APPENDIX B  Additional Resources

Saunders/Mosby/Churchill Livingstone/ Springer-Verlag New York, Inc.

Butterworth-Heinemann www.springer-ny.com
www.us.elsevierhealth.com
Thieme Medical Publishers, Inc.
Schiller Medical SA www.thieme.com
www.schiller.fr
Toshiba America Medical Systems
Section for Magnetic Resonance Technologists www.medical.toshiba.com
(SMRT)
www.ismm.org/smrt Tyco Healthcare
www.mallinckrodt.com
Society for Cardiovascular Magnetic Resonance
(SCMR) University of California, San Francisco
www.scmr.org www.radiology.ucsf.edu

Society of Interventional Radiology (SIR)

www.sir.org
 P R A C T I C E E X A M I N AT I O N S
The following are two practice examinations 3. When exposed to intense static magnetic
of Type A test items (i.e., a stem followed by fields, cultured human cells
four distracters and one answer). Most national a. die prematurely.
examinations and credentialing organizations b. grow more rapidly.
use Type A test items exclusively. c. grow more slowly.
The mix of test items follows that adopted by d. show no effect.
the American Registry of Radiologic Technolo- e. stall in G0.
gists (ARRT) so that each practice examination
contains test items in each category equal to 4. The main potential for biological response
those prescribed by the ARRT. from RF is
The ARRT Content Specifications for the a. carcinogenesis.
Examination in Magnetic Resonance Imaging b. induction of currents.
break down as follows: c. polarization.
d. suppression of relaxation time.
A. Patient Care and MRI Safety 17 questions e. tissue heating.
C. Data Acquisition and 62 questions
Processing 5. Electric current density induced by tran-
D. Physical Principles of 43 questions sient magnetic fields is measured in
Image Formation a. ampere per square centimeter.
Total Test Items 122 questions b. ampere per second.
c. tesla per centimeter.
Specification B, Imaging Procedures, is not d. tesla per second per centimeter.
covered in this textbook and therefore is not e. tesla per second.
covered in these practice examinations.
You should take these examinations one at a 6. The threshold for induction of magnetic
time, separated by additional study of the text- phosphenes at low frequencies is approxi-
book material. The goal should be to score 100% mately
on the second practice examination. a. 1 T/s.
b. 3 T/s.
c. 10 T/s.
MRI EXAM I
d. 30 T/s.
1. When use of an MRI contrast agent causes e. 100 T/s.
tissue to appear brighter, which of the fol-
lowing has occurred? 7. The human responses reported during
a. negative contrast imaging with a 4 T system are presumed to
b. neutral contrast be due to
c. positive contrast a. ferromagnetic projectiles.
d. relaxation contrast b. increased time of exposure.
e. proton density contrast c. the radiofrequency field.
d. the static magnetic field.
2. Relaxation centers are regions of e. transient magnetic fields.
a. increased magnetization.
b. increased proton density.
c. increased susceptibility.
d. reduced T1 relaxation.
e. reduced T2 relaxation.
449
450 Practice Examinations

8. Bone fracture healing is helped by 13. Pulse sequences optimized for imaging the
a. high-frequency RF. CNS are usually not appropriate for body
b. low-frequency RF. imaging because for CNS compared with
c. polarization. body tissues,
d. static magnetic fields. a. PD is too low.
e. transient magnetic fields. b. T1 and T2 are long.
c. T1 and T2 are short.
9. At the start of an MRI examination, patient d. T1 is long and T2 is short.
anxiety is generally due to e. T1 is short and T2 is long.
a. claustrophobia.
b. drowsiness. 14. If a quench occurs and helium escapes, the
c. improper preparation. principal effect on the MRI technologist
d. irritability. will be
e. sleeplessness. a. Donald Duck’s voice.
b. lowered body temperature.
10. Approximately what percent of all patients c. skin rash.
exhibit some claustrophobia? d. sticky hair.
a. 1 e. superficial burn.
b. 5
c. 10 15. Proper quality control for an MRI system
d. 20 requires that test object imaging be per-
e. 50 formed at least
a. daily.
11. Operation of a successful MRI facility must b. weekly.
include c. biweekly.
a. calling each outpatient the day before d. monthly.
the scheduled examination. e. annually.
b. explaining the operating console to the
patient. 16. The principal hazard to patients and per-
c. having the patient show up 1 hour sonnel in an MRI facility is the biological
before examination time. effect of
d. requiring the radiologist to interview a. ferromagnetic projectiles.
the patient before examination. b. the main magnetic field.
e. sending a postcard to remind each c. the RF field.
patient of the scheduled examination d. the sound level.
time. e. the transient magnetic field.

12. Continuing education of MRI technolo- 17. A clearly marked exclusion area should be
gists is identified at the
a. desirable. a. 0.1-mT fringe field.
b. essential. b. 0.3-mT fringe field.
c. required by law. c. 0.5-mT fringe field.
d. unimportant. d. 1-mT fringe field.
e. unnecessary. e. 5-mT fringe field.
 Practice Examinations 451

18. After a 90° RF pulse, the signal received 23. As MXY relaxes to zero,
from the patient is a. gradient magnetic field decreases.
a. a free induction decay. b. nothing happens.
b. a gradient echo. c. signal intensity decreases.
c. a spin echo. d. signal intensity increases.
d. magnetically transferred. e. spin echo is produced.
e. zero.
24. After the relaxation of net magnetization
19. The term pulse sequence refers to to equilibrium, the MR signal
a. gradient magnetic field pulses and RF a. is constant at an intermediate level.
pulses. b. is decreasing.
b. only gradient magnetic field pulses. c. is saturated.
c. only RF pulses. d. is zero.
d. RF pulses and static magnetic field e. oscillates at the Larmor frequency.
pulses.
e. static magnetic field pulses and gradient 25. At equilibrium, if a 90° RF pulse is used,
magnetic field pulses. which of the following MR signals will be
produced?
20. The time-to-echo (TE) is the time between a. free induction decay
the b. gradient echo
a. 180° RF pulse and the next 180° RF c. relaxation signal
pulse. d. spin echo
b. 180° RF pulse and the next 90° RF e. stimulated echo
pulse.
c. 180° RF pulse and the spin echo. 26. Control of TE is exercised by control of
d. 90° RF pulse and the 180° RF pulse. a. proton density.
e. 90° RF pulse and the spin echo. b. T1 relaxation time.
c. T2 relaxation time.
21. The inversion recovery pulse sequence con- d. the 180° RF pulse.
sists of a train of e. the 90° RF pulse.
a. α° … α° … α° …
b. 180° … 90° … 180° … 90° … 27. With the inversion recovery pulse sequence,
c. 90° … 180° … 180° … MR images are constructed from
d. 90° … 180° … 90° … 180° … a. free induction decay.
e. 90° … 90° … 90° … b. gradient echoes.
c. inversion echoes.
22. Which of the following pulse sequences d. spin echoes.
involves all three: TR, TE, and TI? e. stimulated echoes.
a. echo planar
b. gradient echo
c. inversion recovery
d. saturation recovery
e. spin echo
452 Practice Examinations

28. In MRI, the role of the Fourier transform 33. Noise in an image is generally more appar-
is to ent when which of the following are present?
a. change the MR signal into a frequency a. high amplitude signals
spectrum. b. high spatial frequencies
b. measure the amplitude of the MR signal. c. long relaxation times
c. measure the frequency bandwidth of the d. low spatial frequencies
MR signal. e. short relaxation times
d. precisely shape the magnetic field.
e. separate T1 from T2. 34. The raw data of an MR signal contains
what components of the spatial frequency?
29. The Fourier transformation of a spin echo a. angle and frequency
results in data in the b. intensity and frequency
a. amplitude domain. c. magnitude and phase
b. frequency domain. d. phase and angle
c. length domain. e. X and Y coordinates
d. time domain.
e. volume domain. 35. In general, when one samples the data of
the spatial frequency domain with empha-
30. The result of a Fourier transform in MRI sis on the higher spatial frequencies, the
is called a/an a. contrast resolution will be less.
a. amplitude function. b. examination will take longer.
b. image function. c. field of view will be larger.
c. object function. d. signal-to-noise ratio will be less.
d. source function. e. spatial resolution will be less.
e. transform function.
36. The physiologic motion of a nervous action
31. In MRI, smooth objects are represented by potential takes approximately how much
a. a narrow range of frequencies. time?
b. a wide range of frequencies. a. 1 ms
c. high frequencies. b. 200 ms
d. low frequencies. c. 500 ms
e. zero frequency. d. 1 s
e. 5 s
32. The mathematical tool used to analyze the
frequency content of an object is the 37. When an image is constructed, the matrix
a. back projection method. size is equal to the number of different
b. Cartesian number rule. a. B0 values used.
c. Fourier transform. b. frequency-encoding gradients applied.
d. imaginary number rule. c. phase-encoding gradients applied.
e. induction transform. d. RF pulses transmitted into the patient.
e. slice-selection gradients applied.
 Practice Examinations 453

38. The STIR sequences can be designed to 43. The gradient echo pulse sequence can be
suppress the signal from fat because fat identified as:
has a a. α° … α° … α° … α° …
a. high PD. b. 180° … 180° … 180° … 180° …
b. long TE. c. 180° … 90° … 180° … 90° …
c. long TR. d. 90° … 180° … 90° … 180° …
d. short T1. e. 90° … 90° … 90° … 90° …
e. short T2.
44. T2* is shorter than T2, principally because
39. Temporal resolution is a term applied to of
a. tables of temperature. a. differences in T1.
b. temporary resolution. b. differences in T2.
c. the accurate determination of tempera- c. proton density variations.
ture. d. the influence of irreversible magnetic
d. the change in contrast resolution with field inhomogeneities.
body temperature. e. the influence of reversible magnetic field
e. time-related events. inhomogeneities.

40. For a 128 × 128 image matrix to be con- 45. Long repetition times (TRs) result in
structed, which of the following must be a. accelerated T1 relaxation.
applied with 128 different values? b. accelerated T2 relaxation.
a. frequency-encoding gradients c. increased PD.
b. phase-encoding gradients d. longitudinal magnetization that is close
c. proton density amplitudes to equilibrium.
d. RF pulses e. transverse magnetization that is close to
e. slice-selection gradients equilibrium.

41. The FLAIR pulse sequence is a 46. An alpha pulse is a/an

a. Find Late Activity in Recovery. a. pulsed gradient magnetic field.
b. Fluid Attenuated Inversion Recovery. b. pulsed signal reception.
c. STIR sequence with a longer T1. c. RF pulse between 90° and 180°.
d. STIR sequence with a shorter T1. d. RF pulse less than 45°.
e. STIR sequence with high PD. e. RF pulse less than 90°.

42. When a sequence of pulses reduces longi- 47. A pulse sequence diagram should contain
tudinal relaxation by the same amount that all of the following lines of data except
is recovered between pulses, the result is a. MR signal acquired.
a. a saturated state. b. phase-encoding gradient magnetic field.
b. a steady state. c. proton density profile.
c. an equilibrium state. d. slice-selection gradient.
d. T1 amplification. e. transmitted RF pulse.
e. T2 amplification.
454 Practice Examinations

48. For completeness, an MRI pulse sequence 53. For spins to be excited in a given slice of
diagram should contain how many lines of tissue, the RF pulse must
information? a. be omitted.
a. one b. be repeatedly energized.
b. two c. be turned on for a longer time.
c. three d. match the Larmor frequency at B0 plus
d. four BXYZ.
e. five e. match the Larmor frequency at B0.

49. The two properties of any picture element 54. The Q value of an RF pulse is expressed as
are the
a. character and position. a. bandwidth divided by the resonant
b. contrast and spatial. frequency.
c. depth and character. b. resonant frequency divided by the
d. position and size. bandwidth.
e. size and depth. c. resonant frequency times the bandwidth.
d. square of the bandwidth.
50. A pulse sequence is e. square of the resonant frequency.
a. a mathematical algorithm.
b. a spatial rendering of the MR signal. 55. The effect of the pulsed phase-encoding
c. a timeline diagram of MR operation. gradient on the phase of spins along a
d. the name of a controlling subassembly column is
of an MR imaging system. a. nothing.
e. the result of a Fourier transformation. b. to decrease the phase perpendicular to
the gradient.
51. Spatial localization of signals in an MR c. to impress a phase shift along the
imaging system is identified by gradient.
a. B0 intensity. d. to increase the phase perpendicular to
b. collimation. the gradient.
c. filtration. e. to reduce proton density effects along
d. signal encoding. the column.
e. signal enhancement.
56. The gyromagnetic ratio for hydrogen is
52. Regardless of the type of magnet, the Z-axis equal to
is always a. 21 megahertz per tesla.
a. across the patient. b. 21 tesla per megahertz.
b. horizontal. c. 42 megahertz per tesla.
c. parallel with the B0 field. d. 42 tesla per megahertz.
d. parallel with the long axis of the patient. e. 63 tesla per megahertz.
e. vertical.
 Practice Examinations 455

57. Which one of the following RF pulses 60. During a single TR of a partial saturation
should produce the thinnest slice in a 1 T pulse sequence,
imaging system? a. a stimulated echo is formed.
a. 21 ± 0.5 MHz b. an image can be constructed.
b. 42 ± 0.1 MHz c. one gradient echo will be generated.
c. 42 ± 0.5 MHz d. one line of the spatial frequency domain
d. 63 ± 0.1 MHz is produced.
e. 63 ± 0.5 MHz e. one spin echo will be generated.

58. The number of multiple signals that are 61. The purpose of the 90° RF pulse in an
averaged is represented by inversion recovery pulse sequence is to
a. ACQ. a. enhance the proton density.
b. Bug. b. lengthen T1 relaxation.
c. FID. c. lengthen T2 relaxation.
d. SAR. d. lengthen T2* relaxation.
e. SAT. e. rotate the net magnetization onto the
XY plane.
59. Referring to the next figure, which letter in
the pulse sequence represents free induc- 62. What is the time required for double echo
tion decay? imaging compared with that for single echo
a. A imaging?
b. B a. half
c. C b. the same
d. D c. twice
e. E d. four times
e. eight times
A
RFt 63. Which of the following influences the char-
acter of an MRI pixel?
B a. electron density
b. mass density
BSS
c. optical density
C d. proton density
RFs e. slice thickness

64. Which of the following best describes an

D MR image?
Bφ
a. characteristic
E b. dynamic
c. representational
BR
d. static
e. superimposed
456 Practice Examinations

65. Approximately how many colors can the 71. Pixel character is principally determined by
human eye detect? a. B0 intensity.
a. 2 b. extrinsic pressure.
b. 20 c. gradient magnetic fields.
c. 100 d. intrinsic modification.
d. 200 e. RF pulse sequences.
e. 2000
72. The principal mechanism for producing an
66. Which of the following types of image MRI signal is
receptor is used for MRI? a. electromagnetic induction.
a. coil b. motion.
b. film c. proton density.
c. piezoelectric crystal d. spin-lattice relaxation time.
d. scintillation detector e. spin-spin relaxation time.
e. TLD
73. Referring to the next figure, which tissue
67. The ability to detect differences in bright- has the highest equilibrium magnetization?
ness level is termed a. A
a. color perception. b. B
b. conspiquity. c. C
c. contrast perception. d. D
d. definition. e. E
e. visual acuity.
A
68. Pixel location is determined by
a. B0 intensity. B
b. extrinsic pressure. C
c. gradient magnetic fields. D
d. intrinsic modification.
E
e. RF pulse sequences.

69. Abnormal tissue character is best exhibited

by
a. distortion. 1000 2000 3000 4000 5000
b. extrinsic compression.
TR (ms)
c. extrinsic displacement.
d. intrinsic deformity.
e. unexpected pixel brightness. 74. Referring to the figure used in question 73,
at what approximate TR is the longitudinal
70. The principal parameters influencing the magnetization for tissues B and C equal?
character of an MRI pixel are all of the a. 500 ms
following except b. 1500 ms
a. electromagnetic induction. c. 2500 ms
b. motion. d. 3500 ms
c. proton density. e. 4000 ms
d. T1 relaxation.
e. T2 relaxation.
 Practice Examinations 457

75. Referring to the figure used in question 73, 79. A gyroscope wobbles because of an interac-
at a TR of 2000 ms, which tissue should tion between
appear darkest? a. charge and charge.
a. A b. charge and spin.
b. B c. mass and charge.
c. C d. mass and mass.
d. D e. mass and spin.
e. E
80. In the Larmor equation, B is the
76. In a conventional spin echo pulse sequence, a. frequency of precession in megahertz.
after the 90° RF pulse, b. gradient magnetic field.
a. MXY = αM0. c. gyromagnetic ratio in megahertz.
b. MXY = 0. d. magnetic field intensity in megahertz per
c. MXY = M0. tesla.
d. MZ = αM0. e. magnetic field intensity in tesla.
e. MZ = M0.
81. In the absence of an external magnetic
77. Referring to the next figure, which of the field, the direction of a nuclear magnetic
five tissues has the lowest magnetization at moment will
equilibrium? a. be random.
a. A b. be straight down.
b. B c. be straight up.
c. C d. precess.
d. D e. spin.
e. E
82. The gyromagnetic ratio for a given nucleus
a. has a specific value.
A
B b. has units of tesla per megahertz.
C c. is determined by the magnetic field.
D d. varies with B0.
E e. varies with the frequency of precession.

83. When M equals zero, this indicates that

500 1000 1500 2000
nuclear magnetic moments
a. are arranged opposite.
b. are arranged parallel.
c. are arranged randomly.
d. are in motion.
e. have disappeared.
78. Referring to the figure used in question 77,
at approximately 1000 ms, which tissue 84. M0 is proportional to
should appear brightest? a. B0.
a. A b. gyromagnetic ratio.
b. B c. T.
c. C d. T2.
d. D e. the square of the proton density.
e. E
458 Practice Examinations

85. In both the stationary and rotating frame 90. Net magnetization is the result of which
of reference, the Z-axis is always in the property of individual spins?
direction of the a. the difference
a. external magnetic field. b. the motion
b. gradient magnetic field. c. the precession
c. gyromagnetic ratio. d. the relaxation
d. relaxation time. e. the sum
e. proton density.
91. Which of the following increases as M0
86. To rotate net magnetization from the Z- increases?
axis, one must use a. proton charge
a. a gradient magnetic field of proper b. proton density
frequency. c. RF pulse
b. a rotating magnetic field of proper d. T1 relaxation time
frequency. e. T2 relaxation time
c. a static magnetic field of proper
frequency. 92. Which of the following is not a principal
d. the proper T1 relaxation time. MRI parameter?
e. the proper T2 relaxation time. a. Larmor frequency
b. proton density
87. Vector diagrams used to illustrate MRI c. spin density
principles are in which frame? d. T1 relaxation time
a. laboratory e. T2 relaxation time
b. musical score
c. oblique 93. In addition to proton density, signal inten-
d. rotating sity is also affected by
e. stationary a. how the proton is charged.
b. how the proton is chemically bound.
88. In the rotating frame, the net magnetization c. proton valence state.
vector d. the distribution of the proton.
a. decays with T1 relaxation. e. the mass of the proton.
b. decays with T2 relaxation.
c. does not precess. 94. Proton density can best be defined as
d. is the proton density. hydrogen
e. precesses at the Larmor frequency. a. charge.
b. concentration.
89. When tissue is placed in a static magnetic c. configuration.
field, the hydrogen nuclei tend to d. relaxation.
a. align perpendicular to the field. e. translation.
b. align with the field.
c. diffuse. 95. The T1 relaxation time is related to the
d. spin against the field. time required for
e. spin with the field. a. longitudinal saturation.
b. MXY to relax to equilibrium.
c. MZ to relax to equilibrium.
d. translation saturation.
e. transverse saturation.
 Practice Examinations 459

96. Referring to the next figure, which tissue 99. Referring to the next figure, which point
has the highest proton density? represents complete relaxation?
a. A a. A
b. B b. B
c. C c. C
d. D d. D
e. not enough information e. not enough information

A
A
B
C C
D
D
B

100. An FID is a result of relaxation of

a. B0.
97. A given tissue has a T1 relaxation time of b. BXY.
750 ms. When a patient has been out of the c. M0.
magnet for 1.5 seconds, what will be the d. MXY.
value of MZ? e. MZ.
a. 0.14 MXY
b. 0.14 MZ 101. Referring to the next figure, how many
c. 0.14 M0 FIDs are shown?
d. 0.37 MZ a. one
e. 0.37 M0 b. two
c. three
98. Transverse magnetization is symbolized by d. four
a. B0. e. five
b. BZ.
c. M0.
RFt
d. MXY.
e. MZ.

RFs

102. The maximum amplitude of a spin echo

occurs
a. at the beginning of the signal.
b. at the end of the signal.
c. at the midpoint of the signal.
d. variously according to pulse sequence.
e. variously because of motion.
460 Practice Examinations

103. In order to measure MZ, one must use the 109. The term chemical shift relates principally
following pulse sequence: to a change in
a. α° … α° a. Larmor frequency.
b. 180° … 180° b. molecular configuration.
c. 180° … 90° c. nuclear mass.
d. 90° … 180° d. proton density.
e. 90° … 90° e. relaxation time.

104. Magnetization of tissue along the Z-axis 110. A hydrogen NMR spectrum may have
cannot be measured directly because peaks because each nucleus has a slightly
a. MXY is too small. different
b. MZ is too small. a. Larmor frequency.
c. T1 is too short. b. molecular configuration.
d. T2 is too short. c. nuclear mass.
e. the B0 field is not homogeneous. d. proton density.
e. relaxation time.
105. The inversion delay time (TI) is the time
between the 111. A change in resonant frequency for similar
a. α° RF pulse and α° RF pulse. nuclei in a given molecule is caused by
b. 180° RF pulse and 180° RF pulse. a. electron cloud.
c. 180° RF pulse and 90° RF pulse. b. molecular configuration.
d. 90° RF pulse and 180° RF pulse. c. nuclear structure.
e. 90° RF pulse and 90° RF pulse. d. proton density.
e. relaxation times.
106. The term spectrum is also referred to as
a. frequency distribution. 112. Magnetic resonance (MR) signals are con-
b. nuclear species. verted by Fourier transformation from
c. precession. a. Cartesian coordinates to polar coordi-
d. proton density. nates.
e. relaxation time. b. frequency data to phase data.
c. phase data to amplitude data.
107. An NMR spectrum is obtained from a d. spatial frequency domain to spatial
a. free induction decay. location domain.
b. gradient magnetic field. e. spatial location domain to space.
c. precessional state.
d. proton density. 113. A digital image with a 9-bit matrix size
e. relaxation time. would have how many pixels?
a. 64 × 64
108. The Fourier transform of an MR signal b. 128 × 128
results in a distribution of intensity as a c. 256 × 256
function of d. 512 × 512
a. inverse time. e. 1024 × 1024
b. iteration.
c. mass.
d. proton density.
e. time.
 Practice Examinations 461

114. One gigabyte is equal to 119. The chemical shift artifact is usually
a. 25 bytes. a. along the B0.
b. 210 bytes. b. along the RF axis.
c. 215 bytes. c. in the frequency-encoding direction.
d. 220 bytes. d. in the phase-encoding direction.
e. 230 bytes. e. throughout the slice selected.

115. Magnetic resonance imaging (MRI) spatial 120. An image artifact can best be described as
resolution is normally measured in a. an absent anatomy.
a. cycles per centimeter. b. an unwanted pattern that does not rep-
b. cycles per millimeter. resent actual anatomy.
c. line pairs per centimeter. c. positive or negative enhancement of
d. line pairs per millimeter. actual anatomy.
e. shades of gray. d. something absent in the patient.
e. something left behind in the patient.
116. In general, the contrast resolution of an
object will improve as the 121. The partial volume artifact can be reduced
a. high spatial frequencies are adequately by
sampled. a. increasing FOV.
b. image matrix size gets larger. b. increasing TR.
c. low spatial frequencies are rejected. c. obtaining more signals.
d. noise increases. d. reducing flip angle.
e. pixel size gets larger. e. reducing slice thickness.

117. One line pair per millimeter is equal to how 122. The quadrature detection artifact occurs
many line pairs per centimeter? along the
a. 0.01 a. B0 field.
b. 0.1 b. frequency-encoding axis.
c. 1.0 c. longitudinal axis.
d. 5.0 d. phase-encoding axis.
e. 10 e. transverse axis.

118. A 0.5 T MR imaging system is said to have

MRI EXAM II
B0 field homogeneity of ±10 ppm. This is
equal to 1. The use of mineral oil to image the GI tract
a. ±0.5 µT. is an example of enhancing
b. ±5 µT. a. electron density.
c. ±50 µT. b. proton density.
d. ±500 µT. c. T1 relaxation.
e. ±5000 µT. d. T2 relaxation.
e. T2* relaxation.
462 Practice Examinations

2. The use of a paramagnetic agent works by 8. Ferromagnetic surgical clips are hazardous
changing because they
a. electron density. a. can be pulled from the patient.
b. mass density. b. cause signal drop.
c. proton density. c. may heat up excessively.
d. T1 relaxation time. d. may rotate out of the field of view.
e. T2 relaxation time. e. may twist to align with the static mag-
netic field.
3. Transient magnetic fields are measured in
a. tesla per square centimeter. 9. If the presence of foreign ferromagnetic
b. tesla per hertz. material in the patient is questionable,
c. tesla per second per meter. a. CT images should be obtained.
d. tesla per second. b. the examination should be done anyway.
e. tesla. c. fluoroscopy should be done.
d. radiographs may be taken.
4. Exposing pregnant mice to an intense static e. sonograms may be taken.
magnetic field causes
a. congenital abnormalities. 10. A consent form for an MRI examination is
b. growth retardation. necessary
c. malignant disease induction. a. when any imaging system components
d. no effect. are not USFDA approved.
e. reduced maze learning. b. for every examination.
c. when it is the patient’s first MRI
5. Low-frequency electromagnetic fields examination.
(EMFs) relate to d. when the imaging system has been
a. diagnostic ultrasound. repaired recently.
b. microwave radiation. e. when the imaging system is brand new.
c. MRI.
d. typical household electricity. 11. Liquid helium has a vaporization tempera-
e. ultrasonic diathermy. ture of
a. 0° K.
6. The potential hazard to patients from RF b. 4° K.
irradiation is principally due to c. 19° K.
a. current induction. d. 77° K.
b. excitation. e. 100° K.
c. ionization.
d. polarization. 12. A partial saturation image made with a
e. tissue heating. short repetition time is most likely a
a. motion image.
7. It is recommended that the fringe magnetic b. proton density–weighted image.
field be access controlled to c. pure image.
a. 0.1 mT. d. T1 weighted image.
b. 0.5 mT. e. T2 weighted image.
c. 1 mT.
d. 5 mT.
e. 10 mT.
 Practice Examinations 463

13. Routinely, double echo imaging is used to 17. The initial amplitude of an FID is depen-
provide dent on all of the following except
a. magnetic resonance angiograms. a. B0.
b. proton density–weighted images for b. gyromagnetic ratio.
anatomy and T1-weighted images for c. proton density.
pathology. d. relaxation time.
c. proton density–weighted images for e. temperature.
anatomy and T2-weighted images for
pathology. 18. A spin echo pulse sequence consists of the
d. T1-weighted images for anatomy and following train of RF pulses:
T2-weighted images for pathology. a. α° … α° … α° …
e. T2-weighted images for anatomy and b. 180° … 180° … 180° …
T1-weighted images for pathology. c. 180° … 90° … 180° … 90° …
d. 90° … 180° … 90° … 180° …
14. Joint prosthesis in an MRI patient may e. 90° … 90° … 90° …
a. be a contraindication for MRI.
b. be pulled from the patient. 19. In a double echo, spin echo pulse sequence,
c. become ineffective. TE for the second echo is the time from the
d. degrade the image. 90° RF pulse to the
e. heat unacceptably, because of RF a. first spin echo.
exposure. b. next 180° RF pulse.
c. next 90° RF pulse.
15. The pregnant patient d. second 180° RF pulse.
a. can be imaged at any time, after comple- e. second spin echo.
tion of satisfactory consent forms.
b. can be imaged at any time, provided the 20. The inversion delay time (TI) is the time
results will materially affect patient between the
management. a. 180° and the 90° RF pulse.
c. can be imaged after a radiographic b. 180° and the spin echo.
screen. c. 180° RF pulse and the next 180° RF
d. should not be imaged in the first trimes- pulse.
ter with MRI. d. 90° and the 180° RF pulse.
e. should not be imaged with MRI. e. 90° and the spin echo.

16. After a 180° RF pulse, the signal received 21. Of the available net magnetization, the
from the patient is only one that can be observed during an
a. a free induction decay. MR imaging process is
b. a gradient echo. a. BΦ.
c. a spin echo. b. BSS.
d. saturated. c. M0.
e. zero. d. MXY.
e. MZ.
464 Practice Examinations

22. At any point in time, the intensity of the 27. The source function in MRI is a plot of
MR signal is proportional to the size of a. 1/time vs 1/length.
a. BΦ. b. intensity vs 1/time.
b. BSS. c. intensity vs 1/length.
c. M0. d. intensity vs length.
d. MXY. e. intensity vs time.
e. MZ.
28. In the spatial frequency domain, sharp-
23. As MZ relaxes to equilibrium, edged objects
a. an FID is produced. a. approach a single frequency.
b. nothing happens. b. are independent of frequency.
c. signal intensity decreases. c. contain a narrow range of frequencies.
d. signal intensity increases. d. contain high frequencies.
e. spins become saturated. e. have maximum signal amplitude.

24. A free induction decay will be produced by 29. Optimum sampling of an MR signal
which combination of RF pulses? requires that a value be determined
a. α° … 180° a. at least once a cycle.
b. 180° … 180° … b. at least twice a cycle.
c. 180° … 90° c. at least three times a cycle.
d. 90° … 180° d. at least four times a cycle.
e. 90° … 90° … e. more than five times a cycle.

25. A spin echo appears 30. When a representation in the time domain
a. immediately after a 180° RF pulse. is transformed into the frequency domain,
b. immediately after a 90° RF pulse. units are transformed from
c. immediately after an α° RF pulse. a. cm to cm−1.
d. sometime after a 180° RF pulse. b. cm−1 to cm.
e. sometime after a 90° RF pulse. c. hertz to centimeter.
d. hertz to seconds.
26. The time to echo (TE) is e. seconds to hertz.
a. one half the time from the 180° RF
pulse to the next 180° RF pulse. 31. Slice selection during MR imaging requires
b. one half the time of the 90° RF pulse to a gradient magnetic field and a/an
the spin echo. a. broadband RF pulse.
c. the time between 90° and 180° RF b. flat RF pulse.
pulses. c. inverse RF pulse.
d. the time from the 180° RF pulse to the d. shaped RF pulse.
spin echo. e. single frequency RF pulse.
e. the time from the 90° RF pulse to the
spin echo. 32. The unit cycle per millimeter is best matched
to
a. amplitude frequency.
b. Cartesian coordinates.
c. polar coordinates.
d. spatial frequency.
e. temporal frequency.
 Practice Examinations 465

33. Which of the following must be sampled 38. When the Y gradient of a horizontal B0
for a high-resolution image? superconducting magnet is identified as the
a. high spatial frequencies slice-selection gradient, the image plane is
b. high proton densities a. coronal.
c. low spatial frequencies b. irregular.
d. low proton densities c. oblique.
e. short relaxation times d. sagittal.
e. transverse.
34. The middle region of the spatial frequency
domain map is that which best determines 39. Lines of the spatial frequency domain
a. contrast resolution. acquired with weak phase-encoding gradi-
b. proton density information. ents principally contribute information
c. relaxation time data. about
d. signal amplitude. a. contrast resolution.
e. spatial resolution. b. irregular objects.
c. large smooth objects.
35. The different orientation of the net magne- d. small sharp objects.
tization vectors along a row of voxels e. temporal resolution.
represents
a. phase incoherence. 40. In FSE, the zero order spin echo is that
b. proton density. which follows the
c. relaxation time. a. first 180° RF pulse.
d. spatial frequency. b. last 180° RF pulse.
e. temporal frequency. c. lowest amplitude pulse.
d. strongest phase-encoding gradient pulse.
36. Field of view (FOV) relates to the e. weakest phase-encoding gradient pulse.
a. diameter of a pixel.
b. diameter of the area that is recon- 41. When the X gradient magnetic field of a
structed. vertical B0 superconducting magnet is
c. diameter of the patient aperture of the applied as the slice-selection gradient, the
imaging system. image plane is
d. maximum diameter of the patient. a. coronal.
e. size of the imaging coil. b. irregular.
c. oblique.
37. The term trajectory in k-space refers to d. sagittal.
a. each angle and distance. e. transverse.
b. each X and Y point.
c. the method of sampling the amplitude
spectrum.
d. the method of sampling the spatial fre-
quency domain.
e. the method of sampling the temporal
frequency domain.
466 Practice Examinations

42. Lines of the spatial frequency domain 47. At what flip angle does the transverse mag-
obtained with strong phase-encoding gra- netization equal 0.5 M0?
dients are located a. 15°
a. in the center of the spatial frequency b. 30°
domain. c. 45°
b. interleaved throughout the spatial fre- d. 60°
quency domain. e. 75°
c. just outside the spatial frequency
domain. 48. A gradient magnetic field designed to
d. on either side of the spatial frequency inhibit the formation of a stimulated echo
domain. is called a
e. on the periphery of the spatial frequency a. spoiler.
domain. b. steady state.
c. stimulator.
43. Gradient echo imaging is characterized by d. transfer gradient.
a. a refocusing gradient magnetic field. e. truncation.
b. a single 180° RF pulse.
c. a single 90° RF pulse. 49. The character of a pixel element refers to
d. shorter T1 relaxation times. its
e. shorter T2 relaxation times. a. brightness.
b. depth.
44. After a 30° flip angle, transverse magneti- c. location.
zation has a value of d. size.
a. 0.01 M0. e. spectrum.
b. 0.37 M0.
c. 0.5 M0. 50. The two principal timing patterns in a
d. 0.63 M0. pulse sequence diagram are
e. 0.90 M0. a. proton density profile and exposure
time.
45. When compared with T2, T2* b. RF pulses and exposure time.
a. is always longer. c. RF pulses and gradient magnetic fields.
b. is always shorter. d. temporal resolution and contrast resolu-
c. will depend on PD. tion.
d. will depend on T1*. e. temporal resolution and spatial resolu-
e. will depend on T1. tion.

46. Stimulated echoes occur as a consequence 51. In an MR image, a pixel emitting an intense
of MR signal would be rendered
a. equilibrium magnetization. a. black.
b. magnetization steady state. b. bright.
c. magnetization transfer. c. dark gray.
d. T2 relaxation. d. light gray.
e. T2* relaxation. e. void.
 Practice Examinations 467

52. The contrast rendition of an MR image is 56. When two pixels exist in the same magnetic
principally determined by field and one is brighter than the other, it
a. gradient magnetic field amplitude and is probably brighter because of
timing. a. a more intense RF pulse.
b. receiving coil bandwidth. b. a stronger gradient magnetic field.
c. RF pulse amplitude and timing. c. higher proton density.
d. the gray scale resolution of the d. longer imaging time.
computer. e. the postprocessing algorithm.
e. the number of signal acquisitions.
57. The frequency-encoding gradient is
53. The principal control of spatial resolution a. energized during RF excitation.
in an MR imaging system is determined by b. energized during signal acquisition.
a. gradient magnetic field frequency and c. energized after signal acquisition.
timing. d. pulsed at the same time as the phase-
b. proton density range. encoding gradient.
c. RF pulse amplitude and timing. e. that which determines spatial resolu-
d. the gray scale resolution of the tion.
computer.
e. the number of phase-encoding pulses. 58. The selection of a slice of tissue for imaging
requires
54. Gradient magnetic fields serve two princi- a. a gradient magnetic field plus RF
ple purposes: excitation.
a. pixel character and slice selection. b. absence of a gradient magnetic field.
b. pixel intensity and pixel character. c. an oscillation B0.
c. pixel intensity and slice selection. d. a gradient magnetic field and MR signal
d. pixel location within a slice and pixel acquisition.
intensity. e. RF excitation and MR signal acquisi-
e. slice selection and pixel location within tion.
a slice.
59. The “spin warp” method refers to
55. Spins in a coronal slice in a vertical B0 a. a twisted change in frequency after the
superconducting imaging system are selec- gradient pulse.
tively excited when which magnetic field is b. a twisted change in frequency during the
applied? gradient pulse.
a. Bα c. the phase shift impressed along the B0.
b. Bβ d. the phase shift impressed along the
c. BX gradient.
d. BY e. the twisted change in frequency due to
e. BZ the RF pulse.
468 Practice Examinations

60. An MR imaging pulse sequence could not 63. Referring to the next figure, which letter in
work with a single 180° RF pulse because the following pulse sequence represents the
a. T1 relaxation would be too short. frequency-encoding gradient?
b. T2 relaxation would be too short. a. A
c. T2* relaxation would be too short. b. B
d. the resulting magnetization is overpow- c. C
ered by the B0 field. d. D
e. there is no transverse magnetization. e. E

61. Multi-echo imaging in a conventional spin A

RFt
echo pulse sequence results in multiple sets
of images having different B
a. contrast resolution. BSS
b. field of view.
C
c. matrix size.
RFs
d. spatial resolution.
e. temporal resolution. D
Bφ
62. What time is required to produce an image E
having a 256 × 256 matrix from two signal
acquisitions with a repetition time of BR
2000 ms?
a. 17 minutes 64. The MR signal observed in the partial satu-
b. 25 minutes ration pulse sequence is a/an
c. 34 minutes a. conventional spin echo.
d. 50 minutes b. fast spin echo.
e. 8.5 minutes c. FID.
d. gradient echo.
e. GRASE echo.

65. Rods of the human retina

a. are more sensitive than cones.
b. are used primarily for daytime vision.
c. are used principally for photopic vision.
d. respond to color better.
e. respond to intense light levels.

66. The ability to perceive fine detail is called

a. color perception.
b. conspiquity.
c. contrast perception.
d. definition.
e. visual acuity.
 Practice Examinations 469

67. An MR image represents what tissue 72. Referring to the next figure, which diagram
characteristic? best represents T2 relaxation?
a. electron density a. A
b. gyromagnetic ratio b. B
c. hydrogen concentration c. C
d. mass density d. D
e. optical density e. E

68. Pixels in an MR image have two descriptors:

a. character and color.
A B
b. contrast resolution and location.
c. location and character.
d. spatial resolution and contrast resolu-
tion.
e. spatial resolution and location. C D

69. The number of principal MRI parameters

matches the number of primary colors.
That number is
a. one. E
b. three.
c. five.
d. seven. 73. Referring to the next figure, which tissue
e. nine. has the shortest relaxation time?
a. A
70. When an MR image is undergoing postpro- b. B
cessing, ROI stands for c. C
a. a method of observing qualitative d. D
change. e. E
b. a method of obtaining quantitative data.
c. rapid organ imaging.
A
d. region of interest.
e. relaxation on inversion. B
C
71. When the size of an anatomical structure
D
in an MR image is being evaluated, the
appropriate classifications are all of the fol- E
lowing except
a. distorted.
b. intense.
c. large.
1000 2000 3000 4000 5000
d. normal.
TR (ms)
e. small.
470 Practice Examinations

74. Referring to the next figure, at a TR of 77. Liquid nitrogen has a vaporization tem-
300 ms, which tissue should appear perature of
brightest? a. 0° K.
a. A b. 4° K.
b. B c. 20° K.
c. C d. 77° K.
d. D e. 100° K.
e. E
78. Referring to the next figure, the relaxation
A
shown is
a. longitudinal relaxation.
B b. motion relaxation.
C c. proton density relaxation.
D
d. pure image relaxation.
e. spin-spin relaxation.
E

1000 2000 3000 4000 5000

TR (ms)

75. One reason that pure proton density images

(T1 or T2) are not obtained is that it 79. Referring to the next figure, which of the
a. cannot be done. five tissues has the longest relaxation time?
b. costs too much. a. A
c. requires NMR spectrometry. b. B
d. requires system modifications. c. C
e. takes too long. d. D
e. E
76. When considering brain tissue, rank gray
matter (GM), white matter (WM), and A
cerebrospinal fluid (CSF) in increasing B
C
order of net magnetization at equilibrium.
D
a. CSF, WM, GM
b. GM, WM, CSF E
c. GM, CSF, WM
d. WM, CSF, GM 500 1000 1500 2000
e. WM, GM, CSF
 Practice Examinations 471

80. In a partial saturation pulse sequence, 84. Referring to the following vector diagrams,
a. all FIDs are of equal amplitude. which figure properly represents net mag-
b. all spin echoes are of equal amplitude. netization at equilibrium?
c. early FIDs have a higher amplitude. a. A
d. early spin echoes have higher amplitude. b. B
e. magnetization is transferred. c. C
d. D
81. The Larmor equation is best stated as e. E
a. ω = PD × B.
b. ω = B. M0 M0
c. B = PD. B
d. ω = B.
e. B = ω.
B
82. A nuclear magnetic moment in the presence
of an external magnetic field interacts by
a. charge and charge.
b. charge and magnetic field.
B
c. magnetic field and electric field.
d. magnetic field and magnetic field.
e. mass and mass. M0 B M0

83. Net magnetization is defined as the

a. number of nuclei in a patient.
b. number of nuclei in a voxel.
c. sum of nuclear magnetic moments.
d. sum of proton density. B
e. total number of spins.

85. Net magnetization at equilibrium, M0, is

important to MRI because its value
determines
a. precessional frequency.
b. signal intensity.
c. proton density.
d. T1 relaxation time.
e. T2 relaxation time.

86. The laboratory frame of reference is also

called the
a. motion frame.
b. precessing frame.
c. rotating frame.
d. stationary frame.
e. translating frame.
472 Practice Examinations

87. Energy is most efficiently transferred from 92. Strictly on the basis of proton density,
one system to another at which of the following tissues should
a. mass density. appear the darkest?
b. precession. a. cortical bone
c. relaxation. b. fat
d. resonance. c. lung
e. proton density. d. medullary bone
e. muscle
88. Two basic properties of a hydrogen nucleus
important to MRI are 93. Proton density is most closely related to
a. charge and spin. a. bound hydrogen.
b. mass and charge. b. induced hydrogen.
c. mass and concentration. c. mobile hydrogen.
d. precession and mass. d. relaxed hydrogen.
e. spin and magnetic moment. e. transient hydrogen.

89. Because the proton spins, it also 94. The T1 relaxation time is also known as
a. diffuses. a. longitudinal relaxation.
b. has a magnetic moment. b. precession relaxation.
c. induces. c. proton relaxation.
d. precesses. d. translation.
e. relaxes. e. transverse relaxation.

90. Which of the following contributes to the 95. The T1 relaxation time is 300 ms. What
magnitude of M0? will be the value of MZ after a patient has
a. proton charge been in a B0 field for 300 ms?
b. proton density a. 0.37 M0
c. RF pulse b. 0.37 MZ
d. T1 relaxation time c. 0.5 MZ
e. T2 relaxation time d. 0.63 M0
e. 0.63 MZ
91. At equilibrium, no signal can be received
from a patient because 96. The T1 relaxation time for a given tissue is
a. B0 is constant. 600 ms. On removal from a magnet, what
b. M0 is constant. will be the value of MZ after 600 ms?
c. proton density is constant. a. 0.37 M0
d. there is no MXY component. b. 0.37 MZ
e. there is no MZ component. c. 0.5 MZ
d. 0.63 M0
e. 0.63 MZ
 Practice Examinations 473

97. After removal from the magnet for approx- 103. The envelope of an FID is related to
imately five T1 relaxation times, MZ will a. proton density.
equal approximately b. T1*.
a. Zero. c. T1.
b. 0.37 M0. d. T2*.
c. 0.5. e. T2.
d. 0.63 M0.
e. M0. 104. Referring to the next figure, which diagram
can best be used to estimate T1 relaxation
98. Relaxation of transverse magnetization is time?
controlled by a. A
a. motion. b. B
b. precession. c. C
c. proton density. d. D
d. T1 relaxation. e. E
e. T2 relaxation.
A B
99. Relative to T1 relaxation times, T2 relax-
ation times
a. are a little bit longer.
b. are about the same.
c. are just a little shorter.
d. are very much shorter.
C D
e. vary among tissues.

100. The term envelope of an FID refers to the

a. initial proton density.
b. length the signal is straightened.
c. line joining signal peaks.
E
d. motion of the signal.
e. total relaxation time.

101. The FID does not represent true T2 relax- 105. The MZ can only be measured by flipping
ation principally because of magnetization
a. different tissues within the same voxel. a. back to equilibrium.
b. magnetic field inhomogeneity. b. onto the XY plane.
c. magnetic susceptibility effects. c. to the + Z direction axis.
d. motion. d. to the Z direction.
e. proton density inhomogeneity. e. with an alpha pulse.

102. Dephasing of spins within a region contain-

ing the same tissue occurs because of
a. magnetic field inhomogeneity.
b. proton density inhomogeneity.
c. T1 relaxation.
d. T2 relaxation.
e. temperature.
474 Practice Examinations

106. After a 180° RF pulse, 111. The change in resonant frequency among
a. MZ = −M0. similar nuclei in the same molecule is due
b. MZ = 0. to slight changes in
c. MZ = M0. a. proton density.
d. MZ = MXY. b. relaxation times.
e. MZ precesses. c. temperature.
d. the local magnetic field.
107. An NMR spectrum is a graph of intensity e. the transient magnetic field.
as a function of
a. electric potential. 112. The separation of peaks in an NMR spec-
b. frequency. trum increases with increasing
c. mass. a. B0.
d. relaxation time. b. BXYZ.
e. spin state. c. gyromagnetic ratio.
d. relaxation time.
108. The NMR spectrum is obtained from an e. proton density.
MR signal through the process of
a. back projection reconstruction. 113. The number 234 can be expressed in binary
b. complex restoration. fashion as
c. Fourier transformation. a. 01001111.
d. iteration. b. 01010111.
e. signal relaxation. c. 01010111.
d. 01101001.
109. Which nuclear species would show the e. 01101111.
highest signal intensity from an NMR spec-
trum of the human body? 114. One kilobyte is equal to
a. carbon a. 23 bytes.
b. hydrogen b. 26 bytes.
c. nitrogen c. 210 bytes.
d. oxygen d. 215 bytes.
e. phosphorus e. 220 bytes.

110. A nuclear species may exhibit more than 115. The human visual system can resolve
one peak in an NMR spectrum because of approximately how many shades of gray?
its a. 8
a. electron configuration. b. 32
b. mass distribution. c. 64
c. molecular configuration. d. 128
d. nuclear structure. e. 512
e. proton density.
116. The ability to image objects with low
spatial frequency is the ability to image
a. high-contrast objects.
b. low-contrast objects.
c. stationary objects.
d. very large objects.
e. very small objects.
 Practice Examinations 475

117. A high spatial frequency represents 121. The partial volume averaging artifact exists
a. a short signal time. when
b. a very large object. a. a structure is contained within three or
c. good contrast resolution. more slices.
d. poor spatial resolution. b. a structure is not fully contained within
e. rapid changes of the MR signal. a slice.
c. FOV is large.
118. With a 1.0 T imaging system, the soft d. the repetition time is too long.
tissue/fat chemical shift artifact occurs e. the repetition time is too short.
because of a Larmor frequency difference
of 122. To truncate is to
a. 3.5 Hz. a. deform an object.
b. 35 Hz. b. lop off part of the object.
c. 75 Hz. c. reshape an object.
d. 149 Hz. d. slice an object.
e. 300 Hz. e. stretch an object.

119. A 1.0 T MR imaging system is said to have

B0 field homogeneity of ±10 ppm. This is
equal to
a. ±0.1 µT.
b. ±1 µT.
c. ±10 µT.
d. ±100 µT.
e. ±1000 µT.

120. The truncation artifact is more pronounced

when the
a. FOV is large.
b. number of phase-encoding acquisitions
is large.
c. number of phase-encoding acquisitions
is small.
d. repetition time is long.
e. repetition time is short.
 ANSWERS TO CHALLENGE
QUESTIONS

CHAPTER 1 CHAPTER 2
1. Visible light, x-ray, radiofrequency. 1. Except for permanent magnet imaging
2. Radiofrequencies in the range of approxi- systems, both the static and gradient mag-
mately 10 to 200 MHz. netic fields are produced by an electric
3. Felix Bloch, Stanford University (1905- current in a conductor. This is the founda-
1983). Bloch theorized nuclear magnetism tion for electromagnetism.
and proposed equations to explain such a 2. Magnetic resonance imaging rooms are
property. shielded with specially designed conductors
4. Contrast resolution is the ability of an to reduce the intensity of environmental
imaging system to distinguish one soft tissue electromagnetic radiation in the radiofre-
from another. Magnetic resonance imaging quency band.
excels in contrast resolution. 3. Benjamin Franklin first experimented with
5. Spatial resolution deals with high-contrast static electricity in the middle of the eigh-
objects, such as a bone-lung interface, calci- teenth century. Electrostatics is the science of
fied lung nodules, and breast microcalcifica- describing how stationary electric charges
tions. Spatial resolution is the ability to behave.
image very small high-contrast objects. The 4. Both x-rays and radiofrequencies have the
best spatial resolution in medical imaging is same velocity (c = 3 × 108 m/s). X-rays and
x-ray mammography. radiofrequencies are composed of two energy
6. The intrinsic tissue characteristics of proton fields—electric and magnetic—oscillating
density, T1 relaxation, and T2 relaxation are perpendicular to one another. Radiofrequen-
very different for different soft tissues. These cies have lower frequency, longer wave-
differences can be detected and rendered as length, and less energy than x-rays do.
an MR image. 5. Induction is the transfer of energy from one
7. Sensitivity is an indication of how well an state or frame of reference to another without
imaging modality can detect subtle changes touching. In contrast, conduction involves
in anatomy. Specificity is an additional a physical medium in the transfer of such
descriptor relating how well the imaging energy.
system can identify the meaning of those 6. E = hf, where E is energy, h is a physical
differences in diagnosing disease or abnor- constant, termed Planck’s constant, and f is
mal anatomy. frequency. This basically indicates that the
8. f = γB0, where f is the frequency of preces- higher the frequency of electromagnetic
sion, γ is the gyromagnetic ratio (42 MHz/T radiation, the higher is the energy of that
for hydrogen), and B0 is the intensity of the radiation.
static magnetic field. F
7. E = , where F is expressed in newtons and
9. A vector diagram is a graphical relationship Q
illustrating physical quantities that have not Q in coulombs. The newton is the unit of
only magnitude but also direction. kg − m
force and the coulomb is a quantity
10. Precession. A spinning top or gyroscope has s2
mass; when the mass rotates, it generates of electrostatic charge (1 C = 6.24 × 1018
angular momentum. The interaction between electrons).
angular momentum and the gravitational 8. v = λf, where v is the velocity in meters
field results in precession. per second, λ is the wavelength in meters,

476
 Answers to Challenge Questions 477

and f is the frequency of oscillation in the magnitude of alignment is constant and

hertz. results in a net value with the external mag-
9. 1.  Unlike charges attract; like charges repel. netic field. This net value is the equilibrium
2. The force of attraction or repulsion is value.
proportional to the product of the charges 8. M0 is directly proportional to the proton
divided by the square of the distance density, the square of the gyromagnetic ratio,
between them. This is known as Cou- and the intensity of the external magnetic
lomb’s law. field. M0 is inversely proportional to tissue
3. Electrostatic charge is distributed uni-  PDγ 2 B 
temperature  M 0 ≅ .
formly on the surface of a smooth  T 
conductor. 9. Nuclear magnetic resonance spectroscopy is
4. Electrostatic charge is concentrated at conducted at high magnetic fields, up to
regions of irregularity on a nonsmooth approximately 24 tesla, but the most impor-
conductor. tant parameter is the frequency at which the
10. A magnetic field is a force that interacts with NMR signal is received; multiples of 50 and
magnetic poles. The magnetic field is mea- 100 MHz are most often used.
sured in tesla and a magnetic pole in ampere- 10. M = Σµ, where M is the net magnetization,
 1N  Σ is the mathematical symbol meaning to
meter 1T = .
 A − m add, and µ represents each nuclear magnetic
moment.
CHAPTER 3
CHAPTER 4
1. Classical mechanics, often called Newtonian
physics, deals with the interactions of large 1. The gyromagnetic ratio is different for each
objects, such as bowling balls, automobiles, nuclear species. This is a GMIS* parameter
and space ships. Quantum mechanics deals and cannot be changed. *GMIS = God made
with the interactions of very small objects, it so.
such as subatomic particles and electromag- 2. A radiofrequency pulse that is transmitted
netic radiation. The physical laws are differ- into the patient is symbolized as RFt. See
ent for both. Figure 4-11 for a visual representation.
2. The rotating frame of reference. The coordi- 3. The proton spins precess randomly around
nate system is rotating about the Z-axis at the axis of the external magnetic field. They
the Larmor frequency, and the observer is are out of phase, and therefore the sum is
presumed in that frame. projected along the Z-axis.
3. The north and south magnetic poles and 4. The free induction decay (FID) is produced
associated field intensity produced by spin- by the precessing spins that are initially in
ning charged particles. phase but rapidly dephase. The resulting RF
4. The oscillating magnetic field associated signal is at first of high intensity and rapidly
with a 42-MHz radiofrequency. falls to zero.
5. More spins are aligned with the external 5. It is much too weak, and it is not time
magnetic field, and they are in a lower energy varying oscillating and therefore will not
state. electromagnetically induce.
6. MZ and MXY, respectively. 6. Longitudinal magnetization.
7. When a patient is placed in a magnetic 7. MZ = M0, MXY = 0.
field, approximately 1 in 1,000,000 proton 8. The equilibrium magnetization value, M0,
spins align, either with or against this which is principally determined by the mag-
external magnetic field. Once so aligned, nitude of the proton density (PD).
478 Answers to Challenge Questions

9. MZ = 0, MXY = M0. 4. Magnetic field inhomogeneity.

10. The angle through which the net magnetiza- 5. 63%; approximately 5 T1s.
tion vector has been rotated by application 6. Frequency is the rate at which something
of an RF pulse. revolves or spins. Phase is the direction
that multiple spins exhibit at any given
instant.
CHAPTER 5
7. The molecular species in which the hydrogen
1. Because it is emitted immediately after the atom is embedded.
termination of the transmitted RF pulse and 8. That envelope represents the rate at which
the electronics of the receiver must respond the signal relaxes because of T2*. One needs
more quickly. transverse relaxation in a perfectly homoge-
2. 90°-180° … / … 90°-180° … , etc. neous magnetic field for the envelope of the
3. The timing of the energizing of the radiofre- signal to represent T2 relaxation.
quency transmitter and the gradient mag- 9. Tissues with short T1 appear bright, tissues
netic field coils. with long T1 appear dark, but high proton
4. Echo planar imaging (EPI). density enhances the appearance of both.
5. When the repetition time (TR) is not 10. Tissues with long T2 appear brighter than
sufficiently long to allow complete longitu- tissues with short T2; however, all tissues
dinal relaxation of the spin ensemble to with higher proton density will appear
equilibrium. brighter.
6. See Figure 5-2 for an illustration of how the
RF pulse is indicated for a gradient echo
CHAPTER 7
pulse sequence.
7. The time between the initial 90° RF pulse 1. The B0 field is too intense and inhomoge-
and the middle of the spin echo. This time neous.
is twice the time between the 90° RF pulse 2. The saturation recovery pulse sequence with
and the 180° RF pulse. varying repetition time (TR). The inversion
8. See Figure 5-2 for a diagram of the transmit- recovery pulse sequence with varying inver-
ted and received signals for a double echo, sion time (TI).
spin echo pulse sequence. 3. In the middle of the spin echo that was
9. The time between the inverting 180° RF produced.
pulse and the 90° RF pulse. 4. A spin echo.
10. A spin echo. Although the inversion recov- 5. An FID of long duration whose envelope
ery pulse sequence is generally identified as would describe the true T2 of the tissue.
180°-90° … 180°-90° … , in fact, there is an 6. Because all of the net magnetization is along
additional 180° RF pulse that follows each the Z-axis, which is coplanar to the external
90° RF pulse, and that produces the spin magnetic field, B0. Some net magnetization
echo. must project onto the XY plane for a signal
to be detectable.
7. See Figure 7-1.
CHAPTER 6
8. An FID followed by three spin echoes, each
1. Proton density (PD), longitudinal relaxation of decreasing intensity and alternating
time, (T1) and transverse relaxation time polarity.
(T2). 9. There is no signal.
2. T2*, T2, T1. 10. See Figure 7-7 for an illustration of the
3. The concentration of mobile hydrogen, vector diagram that represents tissue magne-
proton density (PD). tization after a 90° RF pulse.
 Answers to Challenge Questions 479

9. 225 Hz.
CHAPTER 8
10. The frequency-encoded axis.
1. A new signal of intensity versus inverse time,
hertz, the NMR spectrum.
CHAPTER 10
2. Aliasing or wraparound artifact.
3. k-Space. 1. The gantry, the operating console, and the
4. At least two data points, samplings, within computer.
each cycle of the MR signal. 2. The operating system.
5. Nyquist. 3. Permanent magnet, electromagnet, and
6. Any bone–soft tissue interfaces, breast mi- superconducting electromagnet.
crocalcifications, and calcified lung nodules. 4. Floating point operation.
7. Yes, the more data acquired, the higher the 5. The gradient coils.
capacity of the computer necessary, and 6. 512 × 512 × 1 = 262,000 bytes or approxi-
the longer it will take to reconstruct an mately one-quarter megabyte.
image. 7. A gas that has been compressed and
8. The gradient coils, which generate the gradi- reduced in volume so that it is now a liquid.
ent magnetic fields. Such a state exists only at a very cold
9. A special form of the Fourier transform par- temperature.
ticularly adapted for the computed genera- 8. Window width, window level.
tion of solutions where the sampled signal is 9. To provide a return path for the magnetic
truncated, allowing the computation to be field and intensify the B0 field.
completed more quickly. 10. To make the B0 magnetic field more uniform
10. See Figure 8-1 for a graphic representation in intensity by the use of shim coils or exter-
of the Fourier transform of a square wave nal ferromagnetic material.
and a spin echo.
CHAPTER 11
CHAPTER 9
1. Electromagnets produce a magnetic field by
1. An NMR spectrum. an electric current conducted through wires
2. Carbon (13C), nitrogen (15N), fluorine (19Fl), fashioned as a solenoid. Permanent magnets
sodium (23Na), and phosphorus (31P). produce a magnetic field by the intrinsic
3. The lines in the spectrum become sharper magnetic property of their material.
and more distinct and more separated. 2. The temperature in the cryostat rises, exceed-
4. Lines that appear in the NMR spectrum ing that of the critical temperature of the
representing the fine detail of spatial rela- cryogen, and the cryogen vaporizes. The
tionship among spins within the same cryogenic gas escapes from the imaging
molecule. system, and the superconducting coils rise in
5. The slight change along the frequency axis temperature so that they can no longer
of the same nuclear species because of the superconduct.
manner in which it is bound within its mol- 3. Both permanent magnets and resistive elec-
ecule. It reflects the magnetic shielding of the tromagnets peak at about 0.3 T. Supercon-
nucleus by the electron configuration. Chem- ducting electromagnets have been produced
ical shift is more obvious at high magnetic with field strengths to approximately 24 T
field strength. but those for imaging are limited by the FDA
6. ±10 ppm. to 3 T.
7. 3.5 ppm, when referred to TMS. 4. 19 K, 5 K, 77 K, 273 K.
8. 42 MHz/T. 5. 21° C, 294 K.
480 Answers to Challenge Questions

6. A precisely machined and positioned pole

CHAPTER 13
face.
7. All electrical, mechanical, molecular, and 1. Its weight and structural loading may dictate
quantum motion ceases. that it not be positioned on an upper floor.
8. It is cooled by water, there is a low-intensity 2. 1.0 mT.
fringe magnetic field, and you can turn it off 3. The absence of a fringe magnetic field.
at night. 4. Use nonmagnetic and nonconducting conduit
9. Insulator—inhibits the flow of electrons. and enclosures. When such conducting con-
Semiconductor—either inhibits or pro- duits are necessary, follow waveguide design
motes the flow of electrons depending on to eliminate extraneous RF. Use DC instead
electrical polarity. Conductor—promotes of AC.
the flow of electrons. Superconductor— 5. The principal advantages are high B0 field
allows the flow of electrons without the loss intensity, uniformity, and signal-to-noise
of energy. ratio. The principal disadvantages are the
10. The ability of an electrical conductor to extensive fringe magnetic field, more patient
promote the flow of electrons without resis- confinement, and higher cost.
tance; there is no loss of energy, and there- 6. To ensure that no patients who undergo
fore no voltage is required. imaging have metallic plants, prostheses, or
surface metal that would interfere with the
quality of the MR image.
CHAPTER 12
7. What would be the effect of the fringe mag-
1. Shim coils are used to adjust the B0 magnetic netic field of the MRI system on nearby elec-
field to its maximum uniformity. For MRI, tronic equipment? What would be the effect
uniformity is termed homogeneity. of nearby iron, stationary or moving, on the
2. Reduced field of view and image nonunifor- quality of the MR image?
mities. 8. A Faraday cage is an electromagnetic shield
3. ±3 µT. constructed of metal sheets or mesh designed
4. The spatial resolution is limited by the pixel to attenuate ambient external RF from the
size, and pixel size is equal to 80 mm ÷ 512 = imaging room.
0.15 mm. 9. Passive shielding uses iron to reduce the
5. 0.025 T/m per 0.1 s = 250 T/m/s. intensity of the fringe magnetic field. Active
6. Reduce the field of view; increase the matrix shielding uses reverse polarity windings of
size. the electromagnet in the cryostat to reduce
7. GX. the intensity of the fringe magnetic field.
8. Because they are closer to the tissue- 10. 3 kHz to 300 GHz.
emitting signal and have a smaller signal
receiving distance, the signal-to-noise ratio
is higher.
CHAPTER 14
9. The read gradient is always frequency
encoded. It can be GX or GY, depending 1. The size of the pixel. Pixel size is determined
on the need to suppress chemical shift by the field of view divided by the matrix
artifacts. size.
10. One that is used to both transmit RF into 2. Both are single numerals, the digit in the
the patient and receive the MR signal decimal number system and the bit in the
from the patient. Normally, head and body binary number system.
coils are homogeneous; surface coils are 3. The amplitude of the phase-encoding
inhomogeneous. gradient.
 Answers to Challenge Questions 481

4. Install gradient coils (GXYZ) to produce gra- signal-to-noise ratio, and therefore the 256
dient magnetic fields (BXYZ). matrix will have better contrast resolution.
5. 212 = 4096 − 4096 individual gray levels are 7. Magnetic resonance imaging is performed
possible with such a system. because of its superior contrast resolution.
6. 2 lp/cm = 0.2 lp/mm = 5 mm/lp∴2.5 mm. Contrast resolution is principally dependent
7. 111011 cm. on the low-amplitude phase-encoding signal
8. 256 × 256 = 65,536 pixels per image. Each acquisitions, which fill the central region of
pixel has 12-bit gray scale, that is, 1.5 bytes k-space. Therefore the central region is more
(212 ÷ 28 = 1.5 bytes). That product is 98,304 important.
bytes or approximately 100,000 bytes = 0.1 8. There is no one-to-one relationship. Each
megabytes × 32 = 3.2 megabytes. area in k-space contains numerical informa-
9. Postprocessing with varying window level tion relating to every pixel in the image.
and width allows the observer to visualize 9. Noise is rather uniform and independent of
the entire 12-bit range. spatial frequency in MRI. Signal increases
10. With increasing spatial frequency, object size with lower spatial frequencies. Consequently,
decreases, which makes it more difficult to larger objects (low spatial frequencies) ex­
resolve both spatially and contrast-wise. hibit better signal-to-noise ratio and there-
Therefore both spatial resolution and con- fore better contrast.
trast resolution are reduced at higher spatial 10. Narrower receiver bandwidth produces
frequencies. higher signal-to-noise ratio and therefore
better contrast resolution.
CHAPTER 15
CHAPTER 16
1. The measure of spatial frequency in MRI is
lp/cm. 1. Five lines, one each for (a) the transmitted
2. k-Space is the spatial frequency domain with radiofrequency pulse and (b) the received
X and Y coordinates. The X coordinates are signal, and one line for each gradient mag-
assigned to frequency, and the Y coordinates netic field.
are assigned to phase. See Figure 15-7. 2. 20 × 2 × 512 = 20.5 seconds.
3. Approximately 10 lp/cm for head and whole 3. The timing of the RF pulses and the type of
body and up to 20 lp/cm for the surface coil. pulse sequence used. Field of view and pixel
4. Trajectory through k-space refers to the size principally influence spatial resolution
manner in which the lines in k-space are but not the contrast or contrast rendition.
filled. The variations of filling k-space are 4. See Figure 18-11 for a diagram of the RF
line by line sequentially, line by line segmen- pulse sequence for an inversion recovery
tally, line by line continuously, and spiral image and the timing for the appearance of
filling. the MR signal.
5. Normally, this will result in a 256 × 256 5. There is no sharp edge to the RF pulse, and
image matrix. Each line represents one MR therefore some bleeding into adjacent slices
signal detected, each signal having been can occur, resulting in image degradation;
sampled and Fourier transformed. It would this is controlled by obtaining slices in a
have been 256 different amplitudes of the noncontiguous fashion.
phase-encoding gradient magnetic field. 1.73
6. 3 = or almost a doubling of signal-to-
6. Spatial resolution improves with smaller 3
pixels, and therefore a 512-image matrix has noise ratio.
better spatial resolution. Contrast resolution 7. At any time that the transmitted RF is
improves with larger pixels because of better energized.
482 Answers to Challenge Questions

8. See Figure 17-20 for an illustration of how 10. Soft tissues have T1 relaxation times mea-
the vector diagram appears when an ensem- sured in hundreds of milliseconds. T2 relax-
ble of spins is partially saturated. ation times are tens of milliseconds.
9. The Z-axis is always drawn parallel to the
B0 magnetic field. Therefore for a horizontal
CHAPTER 18
B0, the Y-axis is vertical, anteroposterior
through the patient, and the X-axis is lateral 1. BSS, the gradient magnetic field, identifies
across the patient. the slice. In a superconducting MRI system
10. Gradient coils are hardware; they conduct having a horizontal B0 field energizing, GZ
electric current and are fabricated to produce will produce a transverse image, GY, a sagit-
a gradient magnetic field with a particular tal image, and GX, a coronal image.
orientation. It is the gradient magnetic field 2. Imaging time is greatly reduced and that
that influences MR signal production and reduces patient motion artifacts.
detection. 3. To restore phase coherence of the spin
ensemble and produce the spin echo for
signal detection.
4. Spin echo fills one line of k-space for each
CHAPTER 17
TR. Fast spin echo fills multiple lines of
1. Visual acuity is the ability to recognize and k-space with each TR by cycling multiple
identify fine detail. Visual acuity is improved 180° RF pulses with varying intensity phase-
when ambient light levels are reduced. Image encoded gradient magnetic fields (BΦ).
masking and low light level illumination in 5. The first Fourier transform is of the MR
the reading room are essential. signal and produces a line in k-space along
2. That will result in a proton density–weighted the frequency-encoded direction. The second
image. Fourier transform is of the orthogonal data
3. The numerical value of each pixel in along the phase-encoded direction of k-space.
MRI represents the value of proton 6. The frequency-encoded gradient magnetic
density, T1 and T2 for the tissue in that field (BR) is shaped in time to accommodate
voxel. moving spins.
4. The early echo image will be T1 weighted 7. Short-term inversion recovery and absolute
and the late echo image, proton density inversion recovery. These are two modifica-
weighted. tions of an inversion recovery pulse sequence.
5. Fluids have long T1 and T2 relaxation times; 8. After multiple RF pulses closely spaced,
therefore, cerebrospinal fluid will appear spins do not fully relax and contribute to a
bright on images with long TR (PDW) and secondary echo. This is the stimulated echo.
long TE (T2W). 9. When data or an object is reproduced as a
6. Magnetic resonance imaging uses the emis- mirror image that is symmetrical about an
sion of electromagnetic radiation. axis, the condition is Hermitian. In k-space,
7. Inversion recovery imaging, in general, pro- Hermitian symmetry is exhibited on either
duces higher-contrast images at the expense side of both axes.
of longer imaging time. 10. Each echo in a train of echoes in FSE is
8. The application of the three gradient mag- generated under a different amplitude phase-
netic fields, BSS, BΦ, and BR, and the fre- encoding gradient (GΦ). The effective echo
quency of the RF excitation pulse. time refers to that echo generated with the
9. Alter the RF pulse sequence timing so that lowest amplitude phase-encoded gradient
the net magnetization of the two tissues are magnetic field (BΦ). This is usually the middle
not the same at the time of sampling. echo.
 Answers to Challenge Questions 483

produce the maximum signal in a given

CHAPTER 19
tissue.
1. Nothing happens. The Larmor frequency at 3. A bipolar read gradient magnetic field, BR.
1.5 T is 63 MHz. At any other frequency no 4. For very short repetition times after a few
interaction occurs. gradient echoes are formed, longitudinal
2. 126 MHz. magnetization reaches a constant level
3. Because it is the most abundant atom in because the degrees of saturation and recov-
the body, 60%, and because it has the ery are equal.
highest gyromagnetic ratio of tissue atoms, 5. The initial negative pole of BR causes rapid
42 MHz/T. dephasing of spins early in the FID. The
4. Magnetization transfer is the transfer of positive pole of BR allows these spins to
tissue magnetization from macromolecules rephase to the level they would have experi-
to water, resulting in improved signal-to- enced and then to dephase again.
noise ratio. 6. See Figure 20-10 for a graphic representa-
5. Hydrogen in water precesses with the fre- tion of the difference between T2 and T2*.
quency 3.5 ppm or 100 Hz lower than fat in 7. B0 magnetic field inhomogeneity. T2* is also
a 1 T MRI system. influenced by chemical shift, tissue magnetic
6. These protons resonate at frequencies susceptibility, and in the case of GRE, the
extending over a wide range depending on read gradient magnetic field.
the molecular configuration. When these 8. Residual transverse magnetization, MXY, is
spins are ignored the result is reduced image spoiled or destroyed after each signal acqui-
contrast. sition in FLASH. Such transverse magnetiza-
7. Increasing receiver bandwidth reduces the tion is rephased in an FISP pulse sequence.
appearance of the chemical shift artifact at all Both are steady state approaches.
B0 field intensities but also reduces the SNR. 9. That flip angle in GRE, which produces
8. 16 kHz ÷ 512 = 31 Hz/pixel; 147 pixels ÷ maximum signal intensity as a function of
31 Hz/pixel = approximately 5 pixels. TR and T1 relaxation.
9. The chemical shift artifact appears along 10. Large flip angles result in large relaxation
the frequency-encoding gradient magnetic and greater signal intensity. However, the
field (BR) for conventional MRI and along process takes longer. Small flip angles
the phase-encoding gradient magnetic field produce less but still sufficient signal inten-
(BΦ) for EPI MRI. sity for good images with a shorter repeti-
10. Use of chemical shift selection with RF tion time and therefore faster imaging.
pulses and dephasing with spoiler gradients
to null the signal from fat. This can also be
CHAPTER 21
done with an inversion recovery imaging
technique that uses an inversion time equal 1. Turbo imaging is gradient echo imaging pre-
to the time when the fat spins are saturated ceded by an 180° RF inversion pulse designed
so that they have no longitudinal magnetiza- to enhance the T1 weighting of the image.
tion to be flipped on the XY plane for signal 2. Gradient and spin echo imaging. This is a
detection (STIR). superfast imaging technique, which acquires
multiple gradient echoes within a single spin
echo.
CHAPTER 20
3. k-Space is filled ten times more rapidly with
1. Faster imaging. turbo technique, but each line of k-space has
2. Gradient echo imaging with a flip angle different T1 weighting because of the initial-
between approximately 30° and 70° to izing 180° RF pulse.
484 Answers to Challenge Questions

4. Offers inversion recovery imaging results in a blip, is applied between echoes so that each
the most RF energy deposition because each of the gradient echoes is acquired under a
line of k-space requires 180°, 90°, 180° RF different phase-encoding gradient (BΦ).
pulse. 6. Because the phase-encoding gradient, BΦ,
5. For turbo imaging, each line of k-space and the read gradient, BR, are cycled so
is acquired with different T1 weighting. rapidly, the transient magnetic field expressed
Because the low-amplitude phase-encoded in T/s is exceptionally large. This can produce
signals contribute most of the contrast, they an unwanted neural stimulation in the
are usually acquired first to fill the central patient.
region of k-space. 7. Spin-spin relaxation (T2). During the time
6. This refers to a high-frequency filter or a of sampling, multiple gradient echoes are
high bandpass filter, which is an electronic formed under the envelope of the spin
term. The receiver electronics allow only echo.
high frequencies to be sampled, reducing the 8. Pulse sequence selection can be used to
low-frequency component of the MR signal. great advantage to accentuate either proton
7. Conventional cardiac MRI is gated to the density or T1 relaxation or T2 relaxation.
cardiac cycle, instead of acquiring one line 9. An additional coil of similar geometry is
of k-space in each cycle. Multiple lines, a placed outside of the gradient coil and
segment of k-space, are filled within each energized with opposite polarity. The result
cardiac cycle. is a reduction in eddy currents, which can
8. See Figure 21-1. be very bothersome in the extremely fast
9. That the phase-encoding gradient is incre- imaging of EPI.
mented for each TR. See Figure 21-3. 10. High B0 field intensity, 1 T or better,
10. Three-dimensional image acquisition where intense gradient magnetic fields, 25 mT/m or
the voxel size has equal dimension for each better, and high slew rate, 100 T/m/s or
of the three sides (e.g., 0.5 mm on a side). better.

CHAPTER 22 CHAPTER 23
1. Echo planar imaging. Images can be obtained 1. Most vascular blood flow is laminar; the
in as little as approximately 50 ms. blood in the center in the lumen is traveling
2. Cardiac MRI and functional brain imaging. faster than the blood near the walls. Plug
3. Echo planar imaging. An entire image can flow exists when the blood moves across the
be produced with a 90° RF pulse followed entire vessel with the same velocity.
by a 180° RF pulse. In gradient echo EPI, 2. Digital subtraction angiography is better
you do not even need the 180° preparation because of the contrast agent filling the
pulse. vessel. Magnetic resonance angiography
4. Because EPI is performed so quickly, chemi- suffers from flow void artifact due to turbu-
cal shift artifacts are large and appear in the lence, but this is rapidly improving with
phase-encoding direction. The inversion contrast-enhanced MRA.
pulse for fat saturation can remove such an 3. Time of flight (TOF) and phase contrast (PC)
artifact. are sensitive to blood flow, and both can be
5. During the formation of the spin echo, a well imaged with maximum intensity projec-
read gradient magnetic field (BR) is cycled, tion (MIP) techniques.
causing the spin echo to be transformed 4. Three-dimensional MRA requires that two
into multiple gradient echoes. An additional gradients act as phase-encoding gradients
small phase-encoding gradient pulse, called but in different directions. The sequence is
 Answers to Challenge Questions 485

repeated for the number of times required by seconds. Sequential images must be obtained
the matrix size. within that time period.
5. Turbulence produces intravoxel dephasing, 9. Exogenous fMRI requires a bolus injection,
which results in loss of signal and produces usually of a gadolinium-DTPA. Imaging
a signal void within the vessel. must commence at the time when the bolus
6. The 3DFT is better for spatial resolution first enters the tissue being imaged.
at the expense of increased imaging time. 10. No effect on any of the three parameters but
However, contrast resolution may be reduced it does shorten T2*.
with thin-slice imaging.
7. 10 mm.
8. Flip angle and repetition time must be care-
CHAPTER 25
fully chosen to maximize the contrast
between blood and stationary tissue. 1. The time constant relating to how quickly
9. T1 weighted imaging, with a short TR. molecules will move from one medium to
10. See Figure 23-1 for a diagram of how the another because of their thermal agitation.
flow rate of blood in the aorta is a function 2. The diffusion coefficient has units of mm2/s,
of time during the cardiac cycle. whereas the b factor is measured in s/mm2.
3. The diffusion coefficient is a property of
tissue compartments. The b factor is influ-
CHAPTER 24
enced by the manner in which gradient mag-
1. Perfusion deals with blood flow in microcap- netic fields are applied and therefore is
illaries; diffusion deals with blood and water dependent on the MRI system.
flow in tissue through the intercapillary 4. Usually, tissue that is rapidly perfused
spaces. appears dark. Tissue that is not perfused
2. BOLD stands for blood oxygen level depen- appears bright.
dent imaging. 5. One of the three gradient magnetic fields is
3. Those sequences that result in perfusion/dif- energized very briefly on either side of the
fusion imaging but particularly diffusion refocusing 180° pulse that produces the spin
imaging. echo.
4. Gadolinium compounds injected intra­ S
6. is signal attenuation. High signal atten-
vascularly are exogenous contrast agents; S0
altered tissue states, such as hemoglobin uation represents increased diffusion.
versus oxyhemoglobin, are endogenous con- S
7. A = = e − bD, where A is the signal atten-
trast agents. S0
5. Deoxyhemoglobin has shorter T2 relaxation uation, S0 is the initial signal strength, and S
time. is the signal strength after application of dif-
6. fMRI requires no ionizing radiation. fusion gradients.
7. Rapid sequence images are obtained, first 8. The signal attenuation is increased exponen-
with oxyhemoglobin and then after a physi- tially. This is a straight-line relationship on
cally agitative task, which consumes oxygen a semilog graph.
causing oxyhemoglobin to change to deoxy- 9. The random movement of particles or mol-
hemoglobin. Subtracting the first image from ecules in a substance due to the temperature
the second results in highlighted areas of of that substance. Brownian motion increases
brain activity. with increasing temperature. At absolute
8. The change following stimulus from the zero, there is no such motion.
oxyhemoglobin to the deoxyhemoglobin 10. Gradient echo imaging and spin echo-echo
state occurs with the time lapse of a few planar imaging.
486 Answers to Challenge Questions

should be nontoxic and concentrate in the

CHAPTER 26
tissue of interest and clear from the body
1. Heart disease, cancer, cerebral vascular rapidly.
disease, and accidents. 2. When body cavities are imaged, such as the
2. Laminar flow appears bright, and turbulence GI tract, such agents greatly disturb the mag-
appears as a signal void. netic field uniformity and therefore reduce
3. X-ray angiography remains the gold stan- T2* and improve contrast.
dard because it has superior spatial resolu- 3. By ingestion, by inhalation, or by intrave-
tion; however, 3DFT MRA is closing in. The nous injection.
2DFT MRA has the poorest spatial resolu- 4. MRI contrast agents generally increase signal
tion of the three. intensity from the affected tissue. This is
4. There is not a simple answer, but MRI is termed positive contrast enhancement.
fast replacing radioisotope imaging because 5. Generally, molecules of the contrast agent
(1) there is little attenuation of signal by cannot cross the blood–brain barrier.
overlying tissue; (2) signal detection in MRI 6. Gadolinium tagged to DTPA. The gadolin-
is much more efficient than that in radioiso- ium is paramagnetic by virtue of seven
tope imaging; and (3) with contrast-enhanced unpaired electrons in various outer shells. It
MRI, both spatial and contrast resolution accelerates dephasing, shortens T2*, and
are superior. reduces signal intensity.
5. Long axis view, short axis view, four- 7. Because the action of both is to accelerate
chamber view, and two-chamber view. dephasing, resulting in shorter T2* in large
6. 1.5 T B0 field intensity; cardiac phased-array doses, signal intensity is reduced. The con-
coil; gradient magnetic field intensity of at trast is negative in nature.
least 25 mT/m with slew rates of at least 8. There is no effect on proton density. Both
100 T/m/s. relaxation times are reduced, causing a loss
7. With cardiac gating instead of a one-line fill of signal on T2-weighted images and an
in k-space during each RR interval, a segment increase in signal on T1-weighted images.
of multiple lines is filled. This allows the 9. Chelation is the action of binding one atom
entire imaging sequence to be completed in or molecule to another. The DTPA serves as
a few heartbeats but reduces temporal reso- a receptor molecule to which the gadolinium
lution of the cine-cardiac study. can be bound.
8. Skip a few heartbeats between signal acqui- 10. The element, gadolinium, is quite toxic by
sition and increase the TR accordingly. itself. Chelating gadolinium to DTPA reduces
9. The HFI stands for half Fourier imaging. In its toxicity significantly.
such a process, only half of the 128 or 256
lines in k-space are filled. The other half are
CHAPTER 28
computed as the mirror image of those filled.
10. Leads should be colinear with the external 1. Any pattern or structure on the image that
magnetic field, B0. They should not be in does not truly represent the anatomy being
contact with the patient’s skin and should imaged.
have no kinks or loops. 2. If the RF transmitter/receiver is not properly
tuned for that particular B0 field, the result
is a noisy image. The noise can be uniform
CHAPTER 27
or wavelike.
1. The contrast agent should improve contrast 3. Increase the slice thickness and/or pixel size.
after a small volume injection, and the level Both of these options will allow more of the
of contrast should be dose dependent. It object to be within one voxel.
 Answers to Challenge Questions 487

4. Ferromagnetic materials have high magnetic MRI. Magnetic resonance imaging is a safe
susceptibility and result in either signal occupation.
dropout or severe image distortion. 3. Deterministic responses exhibit a threshold
5. Misregistration occurs when multiple images exposure. Below that exposure, no response
are added or subtracted as in MRA. Anatomy will occur; above that threshold, the severity
in one image is in an adjacent pixel in a of the response increases with increasing
subsequent image. Patient motion is usually exposure time and intensity.
the cause. 4. Projectiles, under the influence of the B0 field.
6. Water protons and fat protons do not precess 5. Tissue becomes magnetized, and atoms and
at exactly the same Larmor frequency. molecules become polarized. However, this
Therefore the signal received by these effect is exceptionally small; it disappears
two tissues is displaced on the frequency- rapidly on removal from the magnetic field,
encoding axis. Signal enhancement or signal and there are no immediate or lasting physi-
loss can occur. ologic effects.
7. Ghosting occurs when there is a lack of 6. Possible auditory concerns from the loud
phase stability in the MR signal. A common thumping of the gradient coils, potential
source of ghosting occurs when tissue moves reactions to MRI contrast agents, possible
in a periodic fashion, such as the diaphragm mechanical twisting of magnetic surgical
or the CSF; a summation artifact may be clips, interference with pacemaker perfor-
generated and appear as a ghost of the sta- mance, and claustrophobia.
tionary tissue. 7. Gradient magnetic fields change intensity
8. The appearance is just that, a zipper or with time. This property can result in elec-
regular streaklike artifact usually along tromagnetic induction of an electric current.
the frequency-encoding axis. This is due to The current of neurologic pathways can be
external sources of RF being detected by the influenced by such gradient magnetic fields,
receiving coil or undesired MR signals, such causing a variety of physiologic responses.
as those due to stimulated echoes. However, none are life threatening, and all
9. This is caused by aliasing, which results cease when imaging stops.
from undersampling of the MR signal in the 8. Specific absorption rate (SAR) refers to the
phase-encoding direction. Aliasing produces time-related fashion in which RF energy
an image of anatomy of the opposite side of is deposited in tissue. The recommended
the body from where it should appear. Alias- maximum limit is 0.4 W/kg.
ing can be reduced by increasing the field of 9. Heating. Tissue temperature may be ele-
view. vated, and under some circumstances, super-
10. Truncation artifacts are associated with ficial burns are possible.
sharp tissue interfaces. This occurs when the 10. The B0 field is measured in tesla, the gradient
MR signal is so strong that it is outside of magnetic field is measured in mT/m, and the
the sampling window. The appearance is of RF field is measured in hertz. The most
a ringing effect next to that interface. important measure for evaluating biological
response is the specific absorption rate due to
RF irradiation, which is measured in W/kg.
CHAPTER 29
CHAPTER 30
1. Threshold, nonlinear.
2. There is no reason for a pregnant MRI tech- 1. All of the mechanical and electronic appara-
nologist to alter her normal work habits out tus to be used in and around the MRI suite
of concern about the energy fields used with must be nonmagnetic.
488 Answers to Challenge Questions

2. Review the patient’s neurosurgical chart to 7. Have a friend or family member visit with
ensure that no ferromagnetic surgical clips the patient during the examination. Provide
are present. If you are uncertain, a radio- the patient with eyeshades and headphones.
graph may be required to determine if clips Carefully explain the nature of the examina-
are present. The date of the operation tion to such a patient. In some situations,
may help determine if such clips may be patient sedation may be necessary.
ferromagnetic. 8. Accreditation of the MRI facility by the
3. At least 2 years of formal training in medical American College of Radiology is strongly
imaging and certification by the American advised. The cornerstone of this accredita-
Registry of Radiologic Technologists. Addi- tion program is an ongoing quality control
tional formal training and regular continu- program based on daily measurements and
ing education in MRI are also necessary. observations and annual performance evalu-
Certification as an MRI technologist by the ation by medical physicists.
ARRT should be the goal of every MRI 9. Liquid helium and liquid nitrogen have
technologist. vaporization temperatures of 4° K and 77° K,
4. Patients with a cardiac pacemaker represent respectively. During a quench, the gases are
the most sensitive people because the fringe released and can occlude the available
magnetic field may disrupt the operation of oxygen. The system must be vented to the
the pacemaker. The exclusion level is 0.5 mT. outside, and in the event of the quench,
5. A reminder phone call the day before the the patient and personnel should vacate the
examination; help with completing a patient imaging room until the cryogenic gases are
questionnaire; personal consultation about dissipated.
the nature of the examination; and report 10. To inform the patient that there are known
generation in a timely fashion. hazards associated with ferromagnetic
6. To avoid the possible disruption of MRI objects inside the body. This is also a good
system operation because of interference by time to explain that there are no lasting
ferromagnetic objects. Such objects can be effects from the energy fields of MRI.
brought into the imaging room by visitors,
patients, physicians, and other hospital staff.
 ANSWERS TO PRACTICE
E X A M I N AT I O N S
44. e
MRI EXAM I
45. d
1. c 46. e
2. a 47. c
3. d 48. e
4. e 49. a
5. a 50. c
6. b 51. d
7. e 52. c
8. b 53. d
9. a 54. b
10. b 55. c
11. a 56. c
12. b 57. b
13. b 58. a
14. a 59. c
15. a 60. d
16. a 61. e
17. c 62. b
18. a 63. d
19. a 64. c
20. e 65. e
21. b 66. a
22. c 67. c
23. c 68. c
24. d 69. e
25. a 70. a
26. d 71. e
27. d 72. a
28. a 73. a
29. b 74. c
30. c 75. e
31. d 76. c
32. c 77. e
33. b 78. b
34. c 79. d
35. d 80. e
36. b 81. a
37. c 82. a
38. d 83. c
39. e 84. a
40. b 85. a
41. b 86. b
42. b 87. d
43. a 88. c

489
490 Answers to Practice Examinations

89. b 11. b
90. e 12. d
91. b 13. c
92. a 14. d
93. b 15. b
94. b 16. e
95. c 17. d
96. a 18. d
97. c 19. e
98. d 20. a
99. d 21. d
100. d 22. d
101. a 23. d
102. c 24. e
103. c 25. d
104. b 26. e
105. c 27. e
106. a 28. d
107. a 29. b
108. a 30. e
109. a 31. d
110. a 32. d
111. b 33. a
112. d 34. a
113. d 35. a
114. e 36. b
115. c 37. d
116. e 38. a
117. e 39. a
118. b 40. e
119. c 41. d
120. b 42. e
121. e 43. a
122. d 44. c
45. b
46. b
MRI EXAM II
47. b
1. b 48. a
2. d 49. a
3. d 50. c
4. d 51. b
5. d 52. c
6. e 53. e
7. b 54. e
8. e 55. c
9. d 56. c
10. a 57. b
 Answers to Practice Examinations 491

58. a 91. d
59. d 92. a
60. e 93. c
61. a 94. a
62. a 95. d
63. e 96. a
64. c 97. a
65. a 98. e
66. e 99. d
67. c 100. c
68. c 101. b
69. b 102. a
70. d 103. d
71. b 104. d
72. b 105. b
73. e 106. a
74. e 107. b
75. a 108. c
76. e 109. b
77. d 110. c
78. e 111. d
79. a 112. a
80. c 113. c
81. d 114. c
82. d 115. b
83. c 116. d
84. a 117. e
85. b 118. d
86. d 119. c
87. d 120. c
88. a 121. b
89. b 122. b
90. b
 GLOSSARY OF MAGNETIC
RESONANCE IMAGING TERMS

accreditation  The process of being recognized by continuing education  Any of several methods—
an organization, such as CMS (Centers for lecture, course, Web-based, directed reading—
Medicare and Medicaid Services) as acceptable designed to maintain an adequate fund of
to recognize a provider. The Joint Commission knowledge in a particular area.
(TJC), American College of Radiology (ACR), contrast agent  Compound used as an aid for
Intersocietal Accreditation Commission (IAC), imaging internal organs with x-rays.
and RadSite (RS) are current accreditation contrast perception  The ability to visualize differ-
organizations. ences in image brightness levels.
alias  False recognition. Wrap-around artifact. The contrast resolution  Ability of an imaging system
result of inadequate signal sampling. to distinguish adjacent soft tissues from one
amplitude gradient switching  The rate at which another; this is the principal advantage of 
gradient field magnets are turned on and off MRI.
(mT/ms). convolution  The process of converting digital
balanced steady state free precession  SSFP data from one set of numbers to another without
simultaneously acquired in transverse and longi- losing information. Presenting digital data in a
tudinal directions. different format.
blipped echo planar  Rapid excitation of the cryogen  Atmospheric gases, such as nitrogen and
read gradient during a single echo for faster helium, that have been cooled sufficiently to con-
imaging. dense into a liquid.
blood oxygen level dependent  Commonly cryogenic  Relating to liquefied gas at very low
called BOLD. Enhanced MR signal from oxy- temperature such as hydrogen (77° K) and
genated brain tissue. helium (4° K).
blood–brain barrier  This is a protective mem- cryostat  Apparatus for maintaining a constant
brane that prevents harmful substances in the low temperature; requires vacuum chambers to
blood, such as contrast agents, from entering the help with thermal isolation.
brain. decoupled  Disconnecting one component from
bound water protons  Hydrogen bound in water, another temporarily as with the use of surface
not free hydrogen, is the principal source of the coils disconnecting from the body coil during
MR signal. signal acquisition.
breathhold imaging  Patient is instructed to hold deoxyhemoglobin  Venous blood is paramagnetic
breath during fast imaging to reduce motion because it has four unpaired electrons in outer
artifacts. shells.
bytes  Groups of eight bits; represents one charac- dephasing  Loss of phase coherence within an
ter or digit. ensemble of spins.
certification  Program to identify accomplishment dewars  Insulated vessels used to transport cryo-
in a particular field, as in Board Certified Radi- genic gas.
ologist. Similar to accreditation for organiza- diffusion  Process by which molecules or other par-
tions. ticles intermingle and migrate because of the
chelate  Sequestering agent. random thermal motion.
chemical shift  Change in the Larmor frequency of diffusion coefficient  The rate at which molecular
a given nucleus when bound in different sites in components cross a cell membrane or similar
a molecule, owing to the magnetic shielding structure. Symbolized by D (mm2/s).
effects of the electron orbitals. diffusion gradient  Additional gradient magnetic
classical mechanics  The physics of large objects fields impressed before and after the 180° RF
as described by Issac Newton in the seventeenth pulse in spin echo imaging which produces an
century. Complemented by quantum physics of image of membrane integrity.
the twentieth century. digit  A single number.

492
 Glossary of Magnetic Resonance Imaging Terms 493

digital imaging  An image consisting of a pixel transform is used to generate the spectrum 
matrix, each of which has a given numerical from the FID and is essential to most imaging
binomial value. techniques.
downfield  Magnetic resonance spectroscopy (MRS) free induction decay  The signal emitted immedi-
peaks appearing at lower frequency. ately after an excitation pulse; the decay is
dynamic range  Number of possible discrete values caused by progressive dephasing of the spins; the
for each pixel; shades of gray for each pixel (e.g., decay time constant is T2*.
210 = 10 bits = 1024 gray levels). free protons  Different from protons bound to
echo train  Multiple signal echos obtained in a water (most), protons in fat molecules (middle),
single repetition time resulting in faster imaging. and protons bound to other macromolecules
electromagnetic radiation  Oscillating electric (least). Free protons resonate at the highest fre-
and magnetic fields that travel in a vacuum with quency, 47 MHz/T.
the velocity of light. Includes x-rays, gamma frequency distribution  The display of values as a
rays, and some nonionizing radiation (such  spectrum of numbers rather than an average.
as ultraviolet, visible, infrared, and radio  frequency encoding  Use of a gradient magnetic
waves). field to produce a range of frequencies along the
energy sink  A device or place where energy, MR signal to provide information on spatial
usually electrical, is deposited and used, such as position.
a motor, heater, or lamp. frequency synthesizer  An electronic module that
energy source  The device or structure that gener- is tunable and the master frequency source for
ates electrical potential such as a hydroelectric an MR imaging system.
generator. fringe magnetic field  Stray magnetic field that
equilibrium  The state of tissue that is fully mag- exists outside the MR imaging system; the area
netized by a static magnetic field, symbolized as around any magnet having a magnetic field
M0. higher than the magnetic field of earth, which is
Ernst angle  The flip angle that produces the stron- typically between 50 and 100 µT.
gest MR signal at a given repetition time (TR). functional magnetic resonance imaging  Color
It is a function of T1 and TR. images obtained using EPI less than 100 ms to
false contouring  Anatomic image boundaries image brain activity.
appear where there should be none because of gantry  Portion of the computed tomographic or
loss of gray scale resolution. magnetic resonance imaging system that accom-
ferromagnetic material  Substance, such as iron, modates the patient and source or the detector
that has a large positive magnetic susceptibility; assemblies.
it is easily magnetized. G-factor  A descriptor for partial parallel imaging
field of view  Anatomy contained within the in which the normal SNR is divided by the
volume imaged; determined by the product  square root of R, resulting in this fudge factor
of acquisition matrix and pixel size; usually (stands for geometry factor), which gives the
expressed in centimeters. SNR for parallel imaging.
filter  An electronic circuit that passes only selec- gradient magnetic field  Magnetic field that
tive frequencies. changes in intensity in a given direction; a typical
flow-related enhancement  The increase in signal value is 10 to 40 mT/m.
intensity of flowing blood compared with  gyromagnetic ratio (γ)  Ratio of the magnetic
stationary tissue when fully magnetized spins moment to the angular momentum of a particle;
replace saturated spins between RF pulses. this is a constant for a given nucleus (MHz/T).
flow-void  Fast moving blood exits the imaging half Fourier imaging  Image reconstructed from
section before the refocusing RF pulse, resulting an MR data set that fills less than all of k-space;
in reduced or no signal. results in faster imaging but with reduced SNR.
Fourier transform  Mathematical procedure to Hermitian symmetry  The symmetry shown by
separate the frequency and phase components of spatial frequency lines, lines in k-space, on 
a time-varying or spatially varying signal from either side of zero amplitude phase-encoding
the amplitudes as a function of time; the Fourier gradient.
494 Glossary of Magnetic Resonance Imaging Terms

image artifact  False features in an image because low flip angle  A less than 90° flip angle used in
of patient instability (motion) or equipment defi- fast imaging. Also called an alpha angle.
ciencies (improper gradients). magnetic field homogeneity  An absolutely
image reconstruction  Process of changing MR perfect magnetic field of uniform intensity.
signals into an image (e.g., 2DFT, 3DFT). magnetic field inhomogeniety  Variation in mag-
imaginary part  A complex number has two netic field intensity. Expressed as T +/− mT or
dimensions, the real part and the imaginary part, PPM.
both of which are necessary for description (e.g., magnetic moment  Measure of the magnetic
north-south, azimuth, and range). properties of an object or particle (the proton)
informed consent  Patient response to authorizing that causes it to align with the static magnetic
and handling medical information. field.
insulator  Material that inhibits the flow of elec- magnetic susceptibility  Measure of the ability of
trons within a conductor or during heat a substance to become magnetized.
transfer. maximum intensity projection  Computer tech-
intensity-response relationship  The description nique of image reconstruction from a volume
of changing resonant frequency as a function of data matrix; selection of the pixel with the
position along a gradient magnetic field. highest signal intensity along a ray onto a 2D
inversion pulse  The first RF pulse (180°) in an image.
inversion recovery pulse sequence. misregistration artifacts  Artifacts produced when
inversion recovery  Pulse sequence for MRI a mask image and a subsequent image are
wherein the net magnetization is inverted and subtracted.
relaxes to equilibrium with the emission of an moment  Short for magnetic moment. The mag-
MR signal after a 90° RF pulse; a method similar netic field generated by a spinning proton.
to spin echo but each excitation is preceded by multi-echo imaging  Acquiring different images at
an inversion pulse at a time TI. different spin echo times during the same study.
inversion time  Time between the 180° RF inver- multislice imaging  The acquisition of several dif-
sion pulse and the subsequent 90° RF pulse to ferent sections during the time required for one
bring net magnetization onto the XY plane. slice by interleaving the slice-selection gradient.
isocenter  The geometric center position for an myocardial tagging  Cardiac wall motion imaging
MR magnet. using spatially selective presaturation RF pulse
J-coupling  Interaction between two or more techniques.
proton spins on the same molecule through navigator echo  A fast, one-dimensional signal
abnormalities in the electron shells. used to reject images with excessive wall motion.
k-space  Mathematical space in which the Fourier net magnetism  The sum total of all the individual
transform of the image is represented; see also magnetic moments.
spatial frequency domain. nonionizing  Radiation with insufficient energy to
k-space segmentation  Sampling a portion of ionize an atom.
k-space and using that data to reconstruct the nonmagnetic  Material that exhibits no magnetic
entire k-space. properties.
k-space trajectory  The manner in which the entire nuclear magnetism  The magnetic field generated
k-space is sampled (e.g., 2DFT, spiral, square). by a spinning nucleus.
laminar flow  Blood flowing in layers; blood flow null point  The time at which the net magnetiza-
is faster in the center of a vessel and slower near tion crosses the zero axis during IR imaging.
the vessel wall. off-resonance artifact  Increased image noise due
line pair  One bar and its interspace of equal width. to inexact tuning of the RF tuner and/or receiver.
longitudinal relaxation  The relaxation of longi- oxyhemoglobin  Oxygenated hemoglobin result-
tudinal magnetization to the equilibrium value ing from increased brain activity and in increased
after excitation; requires exchange of energy MR signal intensity.
between the nuclear spins and the lattice. parallel imaging  Parallel imaging exploits fea-
Lorentz force  The total force on an electron from tures of phased-array RF coils to significantly
electric and magnetic fields. increase the speed of MR image acquisition.
 Glossary of Magnetic Resonance Imaging Terms 495

paramagnetic  Type of substance with a small but becomes resistive, heat is released, which can
positive magnetic susceptibility; the addition of result in rapid evaporation of liquid helium in
a small amount of paramagnetic substance may the cryostat—a possible hazard that must be
greatly reduce the relaxation times of a tissue; vented.
used as contrast agents in MRI. radiofrequency field  For MRI the RF field
partial saturation  Excitation technique applying extends approximately 10 to 200 MHz.
repeated 90° RF pulses at times on the order of refocusing pulse  The 180° RF pulse used to form
or shorter than T1; partial saturation is also the spin echo from the FID.
commonly referred to as saturation recovery; the relaxation centers  Contrast agents increase the
latter term should properly be reserved for the magnetic field slightly and change the proton
particular case of partial saturation when 90° RF relaxation times, T1 and T2.
pulses are far enough apart in time that the repetition time  The period between the beginning
relaxation of nuclear spins to equilibrium is of a pulse sequence and the beginning of the
complete. succeeding and identical pulse sequence.
parts per million  A measure of molecular content rephase  Causing nuclear spins that are precessing
(PPM). Useful in NMR spectroscopy. randomly to precess in phase and generate an
perfusion  Blood flow through smaller and smaller MR signal.
capillaries. R-factor  The acquisition of low-resolution refer-
phase coherence  Multiple spins all in the same ence images that produce coil sensitivity profiles
position of a periodic cycle; the origin of MR and speed up the imaging by a factor that theo-
signals. retically is as high as the maximum number of
pixel  Acronym for a picture element; the smallest phased array coil elements and RF channels.
discrete part of a digital image display. This factor, R, is also known as the scan “reduc-
plug flow  Uniform blood flow; flow profile in tion factor” or the “acceleration factor.”
which flow in the center of a vessel is the same sampling  The process of interrogating a signal
as that along the walls. at regular intervals for the purpose of
polarized  In the presence of a strong external computation.
magnetic field, proton dipoles become aligned sampling bandwith  The frequency range at which
causing net magnetization as in north-south. an MR signal is recorded.
precession  Gyration of the axis of a spinning body saturated  Following a 90° RF pulse the spin
so as to trace out a cone, caused by the applica- ensemble is described by Mz = 0, Mxy = Mo.
tion of a torque tending to change the direction segmentation  For cardiac cine MRI the number
of the rotation axis. of lines of k-space equals each cine frame.
preventive maintenance  Planned program of semiconductor  Material, such as silicon and ger-
parts replacement at regular intervals. manium, that controls electron flow as a gate by
proton density  Density of resonating proton spins application of a small voltage.
in a given region; one of the principal determi- sensitivity  Ability to detect weak MR signal;
nants of the strength of the MR signal from that ability to image a diseased or abnormal state.
region; hydrogen concentration. sensitivity encoding  A method in which the
pulsatile  Arterial blood flow in forward-backward image is reconstructed from the signals detected
motion. by each coil element in the phased array and then
pulse sequence  Set of RF and/or gradient mag- merged. Known as SENSE.
netic field pulses and time spacings between sequencing system  The electronic organization
them; used to excite spins and spatially encode of RF and gradient magnetic field pulses.
the received signal. shimming  Correction of inhomogeneity of the
quantum mechanics  Physics of extremely small static magnetic field, B0, of an MR imaging
objects, which is based on the concept that all system owing to imperfections in the magnet 
physical quantities can exist only as discrete or to the presence of external ferromagnetic
units. objects.
quench  Loss of superconductivity of the B0 coil signal attenuation  Reduction in the intensity of
that may occur unexpectedly. As the magnet the MR signal.
496 Glossary of Magnetic Resonance Imaging Terms

signal averaging  Method of improving SNR by stimulated echo  The result of additional 180° RF
averaging several FIDs, SEs, or GREs. pulses in FSE. Contributes additional signal in
signal-to-noise ratio (SNR)  Used to describe the FSE.
relative contributions to a detected signal of the superconductor  Substance in which electrical
true signal and random superimposed noise;  resistance essentially disappears at temperatures
the SNR can be improved by averaging several near absolute zero; the superconductor used in
MR signals, by sampling larger volumes, or  MR imaging systems is niobium-titanium.
by increasing the strength of the B0 magnetic superparamagnetic  Contrast agents with more
field. unpaired electrons than paramagnetic agents
site selection  Use of gradient magnetism and resulting in higher magnetic susceptibility and
precise RF frequency to image a given section of regional gradient magnetic field.
anatomy. Susceptibility  See magnetic susceptibility.
slew rate  Time required to switch on or off a temporal frequency  Frequency spectrum or value
gradient magnetic field (T/m/s). represented in time (Hz = cycle/sec).
slice selection gradient  The magnetic field gradi- temporal resolution  Resolving events relative to
ent that is energized to select a section of anatomy time. Compared with spatial resolution, contrast
for imaging. resolution, or energy resolution.
spatial frequency  Measure of resolution; usually threshold intensity  The intensity level of a signal
expressed in line pairs per millimeter (lp/mm). below which it is not recognized.
spatial frequency domain  A dimension of the tissue tagging  Identifying tissue samples by com-
Fourier transform space (like k-space) having bining the sample with a specific molecular
units of inverse distance. agent.
spatial resolution  Ability of an imaging process transient magnetic field  A magnetic field that
to distinguish small adjacent high-contrast struc- changes in intensity with time (mT/ms) or posi-
tures in the object. tion (mT/mm).
specific absorption rate (SAR)  Radiofrequency transverse relaxation  Dephasing of a spin ensem-
energy deposited in the patient; magnetic reso- ble in the XY plane. T2 relaxation.
nance imaging exposure quantity; measured in truncation  The abrupt dismissal of high frequen-
watts per kilogram. cies from an MR signal.
specificity  The ability to precisely identify subtle turbo  One of several fast gradient echo imaging
differences in anatomy. pulse sequences.
spectroscopy  Rendering an MR signal into its turbo factor  In TSE imaging, the number of
component frequency range. profiles per excitation, which is equal to the
spectrum  Array of the intensity of the components number of spin echoes generated after each
of the MR signal according to frequency. excitation.
spin echo (SE)  Reappearance of an MR signal turbulence  Blood flow with a random velocity
after the FID has disappeared; the result of the profile; results in spin dephasing and signal 
effective reversal of the dephasing of the nuclear loss.
spins. upfield  Low frequency resonant peaks in NMR
spin ensemble  The collection of spins, protons, spectroscopy.
that produce an MR signal when excited. visual acuity  The ability to perceive fine detail in
spin warp  Another expression for 2DFT imaging an image.
because the phases of the net magnetization zeugmatography  Term for MRI coined from
vectors twist along the direction of the Greek roots suggesting the role of the gradient
gradient. magnetic field in joining the RF to a desired 
spoiled  The process of dephasing a spin ensemble. local spatial region through nuclear magnetic
spoiler pulse  A gradient magnetic field applied to resonance.
eliminate residual signal by dephasing spins. zipper artifact  Appears like a zipper in an image.
steady state  The condition of constant longitudi- Due to magnetic field inhomogeneities along the
nal magnetism after repeated alpha pulses. frequency- and phase-encoding gradients.
 INDEX
A Artifacts (Continued)
Abdomen, 391f, 400f illustration of, 314f
Absorptiometry, 306 in nuclear magnetic spectroscopy, 314, 314f
Accreditation, 433, 434f-435f classification of, 385b
Accreditation phantom, 434f-435f “cone-head,” , 385, 386f
Acquisition control system, 121 definition of, 384
Action potentials, 214 description of, 84
Active shielding, 136, 137f ferromagnetic materials as, 385, 392
Adenosine diphosphate (ADP), 315 foreign materials as, 385-386, 386f-392f
Adenosine monophosphate, 315 magic angle, 395
Adenosine triphosphate, 315 magnetic susceptibility, 303
Alias, 109 metallic devices as, 388f, 392f
Aliasing misregistration, 399-401
correction of, 177 motion, 339, 397-399, 400f-401f
description of, 106-109, 170, 170f off-resonance, 401
field of view image with, 322, 323f partial volume averaging, 396
in two-dimensional Fourier transform images, 178f source of, 249
Aliasing artifact, 170, 395-396 spark, 188-189
Alpha pulse, 52 surgical clips as, 389f
Alternating current, 23-24, 23f, 34f susceptibility, 83-84, 328, 392t
AM, 37 system-related, 394-397, 401
American Association of Physicists in Medicine, 433 truncation, 396-397, 397f
American National Standards Institute, 414 zero frequency, 397, 398f
American Registry of Radiologic Technologists, 425 zipper, 394-395, 397
AMP. See Adenosine monophosphate Atoms, 26
Ampere, 22 ATP. See Adenosine triphosphate
Amplitude gradient switching, 299 Attraction, 26-27, 27f
Analog-to-digital conversion, 120 Autocalibration, 323-324
Andrew, E. Raymond, 5-6 Autoshimming, 120
Aneurysms, 339
Angiography B
digital subtraction, 343 BΦ, 203
magnetic resonance. See Magnetic resonance B0, 195
angiography B0 field, 46
Angular frequency, 183 Back projection, 175
Anterior chest wall, 387f Balanced steady state free precession, 282-287, 287f
Aortic outflow tract, 289f Bandwidth
Area-of-a-square function, 93 definition of, 188, 198
Arnold-Chiari malformation, 222 narrow receiver, 188f
Arterial capillary blood, 347-348 receiver, 119
Arteriovenous malformations, 339 sampling, 253b, 267f
Artifacts, 188-189 transmitted, 188
aliasing, 170, 395-396 Basal metabolic rate, 415
“bleeding,” , 401, 404f Bentonite, 381
blood flow, 399, 402f Binary number system, 164-165, 164t
body shape, conductivity, and extension as, 393 Biomedical implants, 417-418
chemical shift “Birdcage” resonator, 152, 152f, 155f
description of, 393-394 Bit, 164-166
in echo-planar imaging, 303 “Black-blood cine imaging,” , 365, 370f

Note: Page numbers followed by “f” refer to illustrations; page numbers followed by “t” refer to tables; page numbers
followed by “b” refer to boxes.

497
498 Index

“Bleeding” artifact, 401, 404f Cerebral blood flow, 349

Blipped echo planar, 299, 300f Cerebral blood volume, 350-353
Bloch, Felix, 4-5, 245 Cerebrospinal fluid
Blood flow pixel appearance for, 231t
artifacts caused by, 399, 402f T1 relaxation time of, 229
intensity effects used to measure, 337 T2 relaxation time of, 235f, 236
measurement of, 336-339, 338f Cervical spine, 159f, 262f
types of, 332, 332f Charge
Blood oxygen level dependent imaging, 350, 350f, 361 concentration of, 21
Blood–brain barrier, 382 distribution of, 21, 21f
Body coils, 143-144, 153-154, 155f Charles’ law, 432
BOLD imaging. See Blood oxygen level dependent Chelating, 380
imaging Chemical shift
Bone definition of, 310
cortical, 66 description of, 264-265, 265f
hydrogen nuclei in, 66 nuclear magnetic resonance spectroscopy, 310-311,
Bone–soft tissue interface, 96 311t
Bound water protons, 268 Chemical shift artifacts
Boxcar function, 98 description of, 393-394
BR. See Read gradient in echo-planar imaging, 303
Breathhold imaging, 361, 399, 400f illustration of, 314f
Brown, Robert, 354 in nuclear magnetic spectroscopy, 314, 314f
Brownian motion, 354 Cine magnetic resonance imaging, 363-364
BSS, 203 Cineradiography, 367
Bytes, 165-166 Classical mechanics
BZ coils, 149 definition of, 40
description of, 41-46
net magnetization, 43, 43b
C vector diagrams, 43-45, 47
12
C. See Carbon nuclei Claustrophobia, 416-417, 416f
Capillaries, 345 Clay minerals, 381
Carbon nuclei, 315, 316f Coil(s)
Cardiac cycle, 332, 363, 364f body, 143-144, 153-154, 155f
Cardiac magnetic resonance imaging BZ, 149
advances in, 360 gradient. See Gradient coils
breathhold imaging, 361 head/extremity, 154, 155f
cine, 363-364 inhomogeneous, 153
contrast resolution, 372 matrix, 157-159
field of view, 362 Maxwell, 149
gradient coil capacity, 362 phased array, 157-159, 159f
heart evaluations, 368-373 primary magnetic, 114, 115f
obstacles to, 361 quadrature, 152-153, 153f-154f
pulse sequence, 363-365 radiofrequency, 34, 34f, 59, 115f
radiofrequency coils, 362 cardiac magnetic resonance imaging, 362
spatial resolution, 371 description of, 114-115, 143-144
spin echo, 367 room temperature, 144-146
spiral filling, 368 shim
static magnetic field, 362 description of, 115-116, 115f
system requirements, 361-362 parts per million scale, 144
techniques for, 362-373 surface, 154-157, 156f
Cardiac pacemakers, 417 transmit/receive, 119
Carotid artery plaque, 371 tuning of, 119
Carr-Purcell-Meiboom-Gill pulse sequence, 245-246 X gradient, 149-150, 149f
Cartesian coordinate axis, 8-9 Y gradient, 150, 150f
Cartesian coordinate system, 43-44, 46, 46f, 182 Z gradient, 149, 149f
Center frequency tuning, 119 Coil-dependent noise amplification factor, 325
 Index 499

Comb function, Fourier transform of, 98, 98f Contrast agents (Continued)
Compass, 26-27, 28f route of administration, 380
Complex numbers, 182 signal intensity affected by, 380
Complications spatial frequency and, 188
auditory concerns, 418-419 superparamagnetic, 377-379
biomedical implants, 417-418 tissue perfusion uses of, 382
contrast-related, 418 toxicity of, 379-380
ferromagnetic projectiles, 419, 419b tumor localization and characterization using,
Compton scattering, 37 381-382
Computed tomography types of, 379
abdomen, 391f Contrast enhancement
advantages of, 6 approaches to, 376-380
contrast resolution of, 6t hydrogen content alteration, 377
description of, 194 local magnetic field alterations, 377-379
magnetic resonance imaging versus, 225-226, 225f Contrast perception, 215
pixel brightness in, 195f Contrast resolution
spatial resolution of, 6, 6t cardiac magnetic resonance imaging, 372
Computers definition of, 6
acquisition control system, 121 radiographic, 220
digital signal processors, 122 Convolution, 98-99
functions of, 121-122 Coronary artery angiography, 372
graphics processing units, 122 Cortical bone, 66
image processing, 121-122 Couch, patient, 113-114, 114f, 114t
image reconstruction, 121 Coulomb, 18
objectives of, 121 Coulomb’s law, 20-21, 21b
operating system for, 121 Cryocooling, 138
speed of, 122-123 Cryogen
storage capacity of, 122 description of, 113, 135, 140f
types of, 121-122 replacement of, 432
workstation consoles, 122 Cryogenerator, 140f
Conductors, 24, 25t, 146, 147f Cryostat, 114, 135
“Cone-head” artifacts, 385, 386f CSF. See Cerebrospinal fluid
Cones, 214-215 CT. See Computed tomography
Consent form, 426, 428f
Contiguous slice fast spin echo, 255-256
Continuing education, 425 D
Contrast agents Damadian, Raymond, 5-6
applications of, 381-382 Dark field measurements, 316
biodistribution of, 376f, 380 Decibel levels, 418t
blood–brain barrier passage of, 382 Decoupling, 156
chemical structure of, 379-380 Delayed time to peak, 350-353
classification of, 377b Deoxyhemoglobin, 348, 350
clay minerals, 381 Dephasing
complications related to, 418 description of, 73-74, 81f, 275
description of, 88-89 of free induction decay, 278f
exogenous, 346 of XY magnetization, 53
fast spin echo imaging using, 260-261 Dewars, 142
future of, 382-383 Diamagnetism, 26, 26t
gradient echo imaging, 284-289, 284f-289f Diffusion coefficient, 354, 356
history of, 379 Diffusion gradient, 356
iron, 381-382 Diffusion imaging
magnetic resonance angiography using, 378, 379f, echo-planar imaging, 359
382 pulse sequences, 357-359
paramagnetic, 377-379, 378f, 382 spin echo, 356f
in pregnancy, 440 tissue diffusion, 354-357, 355f
reactions to, 418 Digit, 164
500 Index

Digital computer Electric potential energy, 21-22, 21b

bits, 165-166 Electric power, 24-25, 25b
bytes, 165-166 Electric stimulation, 413
description of, 164 Electrical impedance, 24
quantization, 166-167 Electricity
sampling, 167-170, 168f-169f coulomb, 18
Digital image processing system, 125 definition of, 22
Digital imaging, 162 electric field, 19-20, 20b, 20f
Digital signal acquisition system, 124-125 Franklin’s experiments, 18, 18f
Digital signal processors, 122 Electrification, 19
Digital subtraction angiography, 343 Electrocardiograph electrode placement, 363f
Dipole, 26 Electrodynamics, 22-25
Direct current, 23, 23f, 139 electric current, 22-24, 23f
Direct digital synthesizers, 123 Electromagnet
Discrete Fourier transform, 106, 107f description of, 27, 29f, 32-33, 32f, 132-142
Dose-response relationship, 406, 406f quench, 139-142, 142f
Dots points, 216, 218f resistive, 115-116, 128, 128b, 133-134, 133f, 133t
Double echo steady state, 285, 287f secondary, 115-116
Downfield, 312 superconducting, 134, 134t
Dual echo fast acquisition interleaved spin echo, superconductivity, 134-139, 135f-139f
255-256 superconductor operation, 139-142
Dynamic range, 88, 166-167 Electromagnetic fields, low-frequency, 411
Electromagnetic radiation
characteristics of, 2-3
E description of, 34-37
Earth ground, 19 Maxwell’s field theory of, 28
Echo-planar imaging, 298-302 Maxwell’s wave equation, 35, 35b
advantages of, 302 x-rays and, 4
applications for, 303 Electromagnetic spectrum, 3f, 35-37
blipped, 299, 300f Electromagnetism
cerebral blood flow imaging, 349-350 Faraday’s law, 33-34, 33b
characteristics of, 298 Oersted’s experiment, 30-33, 30f-31f
chemical shift artifact associated with, 303 Electron(s)
cine cardiac magnetic resonance imaging use of, 364 configuration of, 26
concerns regarding, 303 description of, 18, 21f
definition of, 298 Electron beam computed tomography, 302
diffusion imaging using, 359 Electron shielding, 264
hardware requirements for, 302-303 Electrostatic force, 21
history of, 298 Electrostatics
k-space in, 298, 301f, 302 description of, 18-22
multishot, 362 electric potential energy, 21-22, 21b
physiologic stimulation threshold of, 303 laws of, 20-21
pulse sequence for, 299f Encode, 165
speed of, 364 End-diastole, 371
spin echo imaging versus, 256 End expiration breathhold, 361
T2-weighted images using, 301f End-systole, 371
Eddy currents, 148 Endogenous magnetic resonance imaging, 347-353,
Edelstein, William, 5-6, 201 350f-351f
Edison, Thomas, 4 Endorectal coils, 157f
Effective echo time, 258, 258f Energy fields, 406-411
Einstein, Albert, 35-36 combining of, 410-411
Einstein’s equation, 354-356 guidelines for, 414-415, 415f
Ejection fraction, 369-370, 370f human responses to, 411
Electric current, 22-24, 23f nonionizing, 406-407
Electric field, 19-20, 20b, 20f, 409 occupational exposure to, 420
Electric force, 29t radiofrequency fields, 409-410, 414
 Index 501

Energy fields (Continued) FID. See Free induction decay

static magnetic fields, 362 Field of view
description of, 407 aliasing artifact with, 322, 323f
health hazards of, 407 cardiac magnetic resonance imaging, 362
long-term effects of exposure to, 408 definition of, 186
tissue interactions with, 407t pixel size and, 186-187
transient magnetic fields, 408-409, 411-414 receiver bandwidth and, 188
EPI. See Echo-planar imaging spatial resolution affected by, 119
Equilibrium of surface coils, 154-157
net magnetization at, 45b, 50-51, 53-55 5-Deoxyfluoroglucose, 316
nucleus in, 41, 41f Flip angle
Equilibrium magnetization state, 12 description of, 51-52, 60, 272-273
Ernst angle, 279-282, 281f-282f, 288 determination of, 119-120
Exogenous magnetic resonance imaging, 346-347, low, 278-279, 281f, 289
348f optimal, 279-282
of radiofrequency pulse, 292-293
selection of, 118
F transverse steady state and, 284
False contouring, 167 Flow
Faraday cage, 37 intensity effects used to measure, 337
Faraday induction, 34 measurement of, 336-339, 338f
Faraday’s law, 33-34, 33b types of, 332, 332f
Fast Fourier transform, 106, 122 Flow-related enhancement, 334-335, 336f
Fast imaging with steady state precession, 282-284, Flow-void, 334, 335f
283f, 288f Fluid attenuated inversion recovery, 251, 358f
Fast low angle shot, 275-277, 282, 294 Fluorine, 315-316
Fast spin echo imaging FM, 37
clinical applications of, 261 Foramen of Magendie, 399, 401f
concerns in, 259 Foramen of Monro, 399
contiguous slice, 255-256 Foreign materials, as artifacts, 385-386, 386f-392f
contrast enhancement with, 260-261 Forward transform, 185
definition of, 255-256 Fourier, Jean Baptiste Joseph, 91
description of, 253-260 Fourier lines, segmentation of, 293, 293f
effective echo time in, 258, 258f Fourier space, 94-95, 96f
gradient echo imaging and, 295 Fourier transform/transformation, 185-186
J-coupling, 261 of comb function, 98, 98f
k-space ordering in, 259-260 definition of, 93, 185, 306
partial Fourier imaging, 253-255 description of, 14-16, 15f
rapid acquired relaxation enhanced imaging, digital image processing application of, 91-92
255-259, 258f discrete, 106, 107f
stimulated echoes, 261 fast, 106, 122
T2-weighted images, 262f flow and, 105-106, 105f
Fat of free induction decay, 97f
hydrogen nuclei in, 310 frequency domain, 94-97
resonant frequency for, 264 history of, 91
Fat inversion pulse sequence, 266 of image data, 97-103, 98f-103f
Fat protons, 264 imaginary part of, 105, 105f, 182-183, 184f
Fat/water chemical shift, 248b inverse, 93f, 98, 101, 185, 185f, 249
Ferrite magnets, 129 orthogonality of, 97-98
Ferromagnetic materials parts of, 105f
description of, 385, 392, 417, 419 periodicity of, 98
as projectiles, 419, 419b properties of, 93, 93f, 97-98
Ferromagnetism, 26-27, 26t real part of, 105, 105f, 182-183, 184f
Ferrometals, 392f real-space function in, 106
FFT. See Fast Fourier transform of rectangular function, 99f
Fick’s law, 354-356 seperability of, 98
502 Index

Fourier transform/transformation (Continued) Gantry

of source equation, 94 axis orientation in, 196f
spatial localization and, 103-105 description of, 112
three-dimensional, 175, 177 permanent magnet imaging system, 116, 116f
two-dimensional, 99, 122, 174 resistive electromagnet imaging system, 115-116
uses of, 306 superconducting magnetic resonance imaging system,
FOV. See Field of view 113-115, 113f-114f
Fovea centralis, 215 Gauss, 29
Frame of reference. See Reference frames Gauss’s law, 28, 28b
Franklin, Benjamin, 18, 18f Geometry factor, 325-327
Free induction decay, 11-14, 12f, 14f, 54-56 Ghosting, 397
amplitude of, 88f, 232 Gradient coils
continuous, 168f capacity of, 362
dephasing of, 278f casing of, 148
envelope of, 74f combined gradients, 150-151, 150f
Fourier transform of, 97f conductors, 146, 147f
with 90° radiofrequency pulse, 62 eddy currents controlled using, 148
nuclear magnetic resonance spectrum from, 306, function of, 196
309f illustration of, 115f
repetition time and, 229 Lorentz forces, 148
spatial frequency of, 199-200 magnetic fields produced by, 195
spin echo and, 80 positioning of, 146f
in steady state, 286f power supply for, 123-124
suppression of, 286f self-shielding, 148-149
Free induction decay line, 397, 399f sets of, 146, 146f
Free protons, 268 shimming using, 145
Free water protons, 264 switching of, 147-148
Frequency X, 149-150, 149f
angular, 183 Y, 150, 150f
spatial. See Spatial frequency Z, 149, 149f
Frequency distribution, 306 Gradient echo images, 276f
Frequency domain, 94-97 chemical shift artifact on, 394
Frequency encoding, 248 series of, 87f
Frequency-encoding gradient magnetic field, 176, 200, spin echo used to generate, 274-275
202-204, 248f, 274f Gradient echo imaging
Frequency synthesizer, 123, 151 acronyms used in, 245t
Fringe magnetic fields, 128, 131 contrast-enhanced techniques, 284-289, 284f-289f
FT. See Fourier transform Ernst angle, 279-282, 281f-282f, 288
Full Fourier encoding, 328, 328f fast low angle shot, 275-277, 282, 294
Function, 92 fast spin echo imaging and, 295
convolution of, 98-99 inversion pulse, 290-291, 291f
sinc, 93, 100f, 101, 185 liver, 292f
Functional magnetic resonance imaging, 345, 349 low flip angle, 278-279, 281f
poiled, 277
pulse sequences, 273, 273f, 285
G refocused, 282-284
g-factor, 325, 326f signal generation in, 274
Gadodiamide, 379 spoiled, 282-284, 283f
Gadolinium-DTPA steady state, 275-278, 280f
chemical structure of, 380f T2 versus T2*, 274-282
echo-planar imaging using, 371 turbo imaging versus, 292
placental transfer of, 418 Gradient magnetic fields
uses of, 377, 378f description of, 143
Gadopentetate dimeglumine, 361 flow compensation, 336
Gadoterate meglumine, 379 frequency-encoding, 176, 200, 202-204, 248f, 274f
Gadoteridol, 379 generation of, 146
 Index 503

Gradient magnetic fields (Continued) Hounsfield units, 167, 195

intensity of, 198 Hutchinson, James, 5-6, 201
phase-encoding, 177, 200-202, 247-248, 248f, Hydrocephalus, 222f
253-254 Hydrogen, 8
proton resonant frequencies, 197f in bone, 66
purpose of, 196 gyromagnetic ratio for, 263
read, 273-274, 274f oxygen binding to, 264
slope of, 198 proton resonant frequency, 263, 264t
spatial localization created using, 175 signal amplitude affected by, 66
X, 200f Hydrogen nuclei, 313-314
Y, 201f mobile, 66, 67t
Gradient moment nulling, 336 precession of, 263
Gradient pulses, 148f
Graphics processing units, 122
GRAPPA, 320, 324, 325f I
Gravitational field, 20 Image. See also specific image
Gravitational force, 29t data, Fourier transform of, 97-103, 98f-103f
Ground, 19 definition of, 213-216
Gyromagnetic ratio, 11, 11t, 42, 50, 306-307 description of, 216
Gyroscope, 9-10, 9f evaluation criteria for, 221-225, 222f-224f
gray scale resolutions of, 168f
inversion recovery, 233-234, 234f
H medical, 213-214
Half Fourier acquired single shot turbo spin echo, pattern recognition for, 216, 217f
254-255, 257f pixel character of, 222-223
Half Fourier imaging, 254, 256f-257f, 366-367 quality of, 205-206
Hard radiofrequency pulse, 51-52 radiographic, 218-221, 220f
Hazards reconstruction of, computer for, 121
auditory-related, 418-419 representational, 217-218
biomedical implants, 417-418 spatial criteria of, 221-223
contrast-related, 418 visual, 214, 215f
ferromagnetic projectiles, 419, 419b Image contrast, 76-77
Head/extremity coils, 154, 155f Image matrix, 172-173, 172f
Hearing loss, 418-419 Image processing, 121-122
Heart Imaging time, 205b
anatomy of, 369-371, 369f Imaging window, 36
cardiac magnetic resonance imaging of, 368-373 Impedance, 24
cine magnetic resonance imaging of, 363-364 Implants, 417-418
left ventricular outflow tract, 370 Induction, 19
myocardial perfusion, 371-372 Infusion pumps, 125
short axis views of, 294f Inhomogeneities, magnetic field, 79
tissue of, 360-361 Inhomogeneous coils, 153
two-dimensional image of, 289f Inorganic phosphate, 315
Heart disease, 360 Insulators, 24, 25t
Heart rate, 362 Intensity-response relationship, 407, 407f
HeLa cells, 410 International Radiation Protection Association, 415,
Helium, 432 415t
Helium recirculation system, 141f Inverse Fourier transform, 93f, 98, 101, 185, 185f, 249
Hermitian symmetry, 254 Inverse transform, 92b, 185, 189f
Hertz, 313 Inversion-delay-time, 62
Hertz, Heinrich, 36 Inversion pulse, 290-291, 291f
Hinsaw, Waldo, 5-6 Inversion recovery
Homogeneity, 144 fluid attenuated, 251
Horizontal spatial frequencies, 174 short time, 251
Horseshoe magnets, 42, 42f weighted images, 232-234, 234f
Hoult, David, 5-6 Inversion recovery imaging, 250-253
504 Index

Inversion recovery magnitude, 254f Liver (Continued)

Inversion recovery pulse sequence, 62, 62f, 87-88, 89f, gradient echo imaging of, 292f
207-208, 207f, 232-233, 233f T2-weighted fast spin echo of, 262f
Inversion recovery spin echo pulse sequence, 252f Longitudinal magnetization, 277-278, 291
Inversion time, 86-87, 88f Longitudinal relaxation, 53-54
Ionizing radiation, dose-response relationship to, 406, Longitudinal relaxation time, 68
406f Lorentz force, 115, 148
Iron, 381-382 Lorentzian force, 31, 31b
Iron oxide particles, 377 Low flip angle, 278-279, 281f, 289
Iron yoke, 116, 116f, 130-131 Low-frequency electromagnetic fields, 411
Isocenter, 14, 114

J M
J-coupling Macromolecular protons, 266
description of, 261 Magic angle artifacts, 395
fast spin echo imaging and, 261 Magnavist, 418
nuclear magnetic resonance spectroscopy, 312-313, Magnet(s)
313f description of, 8
Joint prostheses, 440 electromagnet, 27, 29f, 32-33, 32f
ferrite, 129
K horseshoe, 42, 42f
isocenter of, 14, 114
k-space permanent, 128-132, 129f-132f, 132t
description of, 177, 181 personal belongings not allowed near, 437b
in echo-planar imaging, 298, 301f, 302 power supply for, 124
filling of, 293f, 366-368, 368f superconducting, 113-115, 113f-114f
ordering of, 259-260, 292 types of, 112-113
parallel imaging and, 320, 323-324 visitors’ effect on, 436-437
quadrants of, 366-367, 366f Magnetic bricks, 130, 130f
sampling of, 191 Magnetic coils
segmented/segmentation of, 191, 363, 367-368, 367f primary, 114, 115f
in spin echo imaging, 249f, 254 secondary, 114-115
spiral filling of, 368 Magnetic domains, 25-26, 25f, 26t
symmetrical, 254 Magnetic field(s), 29, 29b
k-space trajectory, 300 description of, 17
Kaolin, 381 fringe, 128, 131
Kilobyte, 165 homogeneity of, 132
Kiloelectron peak, 4 inhomogeneities, 79, 82, 390f
Kilovolt peak, 4, 7 static, 362
description of, 407
L health hazards of, 407
Laminar flow, 332, 332f, 337f long-term effects of exposure to, 408
Larmor equation, 10-11, 11b, 42, 42b, 175, 196 transient, 408-409, 411-414
Larmor frequency, 11-12, 11b, 13f, 43, 46-47, 50-51, Magnetic field dependence, 311, 311f
53, 144, 306, 386 Magnetic field drift, 124
Late spin echo, 240f Magnetic force, 28-29, 29t, 31b
Lauterbur, Paul, 5-6, 16 Magnetic lines of induction, 27
Left ventricular outflow tract, 370 Magnetic moment, 5
Lenz’s law, 33 Magnetic permeability, 27
Lightning, 19, 19f Magnetic resonance angiography
Line pairs, 181, 181f aneurysms on, 339
Liquid helium, 432 applications of, 339
Liquid nitrogen, 137-138, 432 arteriovenous malformations on, 339
Liver blood flow on, 332, 332f
contrast-enhanced magnetic resonance imaging of, contrast-enhanced, 378, 379f, 382
381f coronary arteries, 372-373
 Index 505

Magnetic resonance angiography (Continued) Magnetic resonance imaging computers

description of, 243 acquisition control system, 121
digital subtraction angiography versus, 343 digital signal processors, 122
flow measurement in, 336-339, 338f functions of, 121-122
flow-related enhancement, 334-335, 336f graphics processing units, 122
flow-void, 334, 335f image processing, 121-122
magnetization transfer techniques in, 268 image reconstruction, 121
magnitude effects, 333-335, 333f-335f objectives of, 121
maximum intensity projection, 341, 342f-343f operating system for, 121
mechanism of action, 332-333 speed of, 122-123
phase contrast, 340, 356, 356f storage capacity of, 122
phase shift effects, 335, 338f types of, 121-122
T1-weighted images for, 382 workstation consoles, 122
three-dimensional, 341-343, 342f-343f, Magnetic resonance imaging parameters
372-373 description of, 226
time of flight, 339-340, 372 overview of, 65
turbulence, 333, 334f proton density, 66, 226-227, 227f
two-dimensional, 340, 341f T1 relaxation time, 66-69, 67f-68f, 68t, 74
Magnetic resonance image. See also Image T1p relaxation time, 68-69, 69f, 74
character of, 225-227 T2 relaxation time, 69-74
definition of, 221 phase coherence, 71-72
description of, 221 summary of, 74
evaluation criteria for, 221-225, 222f-224f T1 relaxation versus, 71
formation of, 194 in tissue, 70t
matrix sizes of, 165 Magnetic resonance imaging system
methods of improving, 205-206 accreditation phantom for, 434f-435f
pure, 226-227 ancillary equipment for, 423-424, 423b
Magnetic resonance imaging components of, 112, 112f
advantages of, 6-8 distributed computation on, 121
complex numbers in, 182 electronics
computed tomography versus, 225-226, 225f ancillary equipment, 125
contraindicated devices with, 418b description of, 112, 123
contrast resolution of, 6, 6t digital image processing system, 125
free induction decay, 11-14, 12f, 14f digital signal acquisition system, 124-125
hardware for, 8 frequency synthesizer, 123
history of, 2-6 gradient coil power supply, 123-124
image contrast, 376 magnet power supply, 124
image creation, 4 radio frequency amplifier, 123
Larmor equation, 10-11, 11b sequencing system, 124
motion in, 332 features of, 129t
multiplanar imaging, 7 gantry
net magnetization, 8-9, 8f-9f, 11, 12f description of, 112
no ionizing radiation for, 7-8 permanent magnet imaging system, 116, 116f
overview of, 8-16 resistive electromagnet imaging system,
parameters for, 7 115-116
precession, 9-10, 9f superconducting magnetic resonance imaging
projection reconstruction, 175, 176f system, 113-115, 113f-114f
protocols, 429-430, 430t maintenance of
radiofrequency pulses in, 52 cryogen replacement, 432
request form for, 426, 427f magnet quench, 432-433
scheduling of, 425-429 overview of, 430
signal, 162-163 preventive, 430-431
spatial resolution of, 6-7, 6t nonmagnetic equipment with, 424
three-dimensional Fourier transform, 177 nuclear magnetic resonance spectrometer versus,
tissue appearance of, 376 175
two-dimensional Fourier transform, 176-177 open, 129
506 Index

Magnetic resonance imaging system (Continued) Megabyte, 165

operating console, 116-120 Megahertz per tesla, 11
description of, 112, 116-117 Meniscal tear, 289f
illustration of, 117f Metallic stent artifact, 388f
image acquisition, 118-119 Microwave radiation, 410
image processing, display, and manipulation, 120 Middle ear prostheses, 440
image scanning, 120 Milliampere-second, 7
prescan calibrations, 119-120 Millitesla, 420
start-up, 117-118 Misregistration artifacts, 399-401
permanent, 129f Mitral valve, 289f
personnel with, 439 Mobile hydrogen nuclei, 66, 67t
quality control for, 433-435, 433t, 434f-435f, 435t Modulation transfer function, 91-92, 182, 182f
repair of, 431 Moment, 336
safety of, 435-439 Motion, 332
staffing of, 424-425, 424t Motion artifacts, 339, 397-399, 400f-401f
warranty period for, 430 suppression technique for, 336
zoning of area for, 437-439, 438f types of, 339
Magnetic resonance imaging window, 37 Motion effect, 177, 178f
Magnetic resonance medical director, 439-440 Motion reduction, 377
Magnetic resonance safety officer, 439-440 MRA. See Magnetic resonance angiography
Magnetic resonance spectroscopy, 7. See also Nuclear Mu, 41-42
magnetic resonance spectroscopy Multi-echo spin echo, 60-62, 210, 211f
Magnetic resonance spectrum, 265f, 266 Multiplanar imaging, 7
Magnetic susceptibility Multiple sclerosis, 225f
artifacts caused by, 83-84, 303, 328 Multislice, multi-echo spin echo imaging, 210,
description of, 392 211f
Magnetism Multislice imaging
laws of, 26-29 definition of, 206
magnetic domains, 25-26, 25f, 26t description of, 206, 208-209
overview of, 25-29 pulse sequence for, 210f
Magnetization Multislice spin echo imaging, 364-365
longitudinal, 277-278, 291 MXY, 45, 50, 52, 66
net, 77 dephasing of, 74
control of, 45-46 description of, 323
description of, 8-9, 8f-9f, 11, 12f, 43, 43b loss of, 72f
at equilibrium, 45b, 50-51, 53-55 relaxation of, 70, 71f
flipping of, 45-46 MXY precession, 53, 54f
T1-weighted images, 85 Myocardial mass, 371
preparation schemes for, 119 Myocardial perfusion, 371-372
transverse, 45, 247-248, 292-293, 300-302 Myocardial tagging, 361
Z-axis, 292-294 MZ, 45
Magnetization prepared rapid gradient echo technique, changes in, 66-67
294, 295f relaxation of, 55f, 67f
Magnetization transfer, 266-268, 266f-268f
Magnetization vectors, 72-73
Mansfield, Peter, 5 N
15
Matrix coils, 157-159 N, 316-318
23
Matter, states of, 135b Na, 315
Maximum intensity projection, 341, 342f-343f Narrow receiver bandwidth, 188f
Maxwell coils, 149 Navigator echo, 361
Maxwell’s field theory of electromagnetic radiation, 28 Net magnetization, 77
Maxwell’s wave equation, 35, 35b control of, 45-46
Medical images description of, 8-9, 8f-9f, 11, 12f, 43, 43b
description of, 213-214 at equilibrium, 45b, 50-51, 53-55
paintings versus, 216 flipping of, 45-46
representational nature of, 217-218 T1-weighted images, 85
 Index 507

Net magnetization vector Operating console (Continued)

description of, 51, 53, 53f, 182-183 image acquisition, 118-119
XY plane component of, 183, 183f image processing, display, and manipulation, 120
Net transverse magnetization, 253 image scanning, 120
90° RF pulse, 48, 52, 56, 61f, 86, 229, 248 prescan calibrations, 119-120
Niobium-titanium alloys, 135, 137, 139f start-up, 117-118
Nitrogen, 316-318 Optical receptors, 214-216
Nonionizing energy fields, 406-407 Orthogonality, 97-98
Nonmagnetic equipment, 424 Oscillator, 4
Nuclear magnetic moment, 40, 41f-42f Oxyhemoglobin, 346f, 348, 350
Nuclear magnetic resonance, 4, 39
Nuclear magnetic resonance imaging, 4
Nuclear magnetic resonance spectrometer, 175 P
Nuclear magnetic resonance spectroscopy Pacemakers, 417
chemical shift in, 310-311, 311t, 314, 314f Parallel imaging
history of, 5, 245 applications of, 327-329, 328f-329f
hydrogen, 310f approaches to, 320
J-coupling, 312-313, 313f B0 field strengths, 327
magnetic field dependence, 311, 311f definition of, 319-320
magnetic field strengths of, 43 description of, 320-321
nuclear species, 306-310 elements of, 320
nuclei, 313-318 extensions of, 330
overview of, 305-306 geometry factor, 325-327
parts per million scale, 199b, 311-312 GRAPPA, 320, 324, 325f
signal intensity, 312, 312f image-based reconstruction, 321-323
Nuclear magnetic resonance spectrum, 14, 308f imaging time, 320
Nuclear magnetism, 4 k-space based, 320, 323-324
Nuclei SENSE, 321-322, 324, 325f
carbon, 315, 316f signal-to-noise ratio, 325-327, 326f-327f
fluorine, 315-316 susceptibility artifact, 328
hydrogen, 313-314 x-ray computed tomography versus, 322f
nitrogen, 316-318 Paramagnetic contrast agents, 377-379, 378f, 382
phosphorus, 314-315, 315f Paramagnetism, 26, 26t
sodium, 315, 316f “Partial flip” pulse, 52
Nucleus Partial flip pulse sequence, 60
in equilibrium, 41, 41f Partial Fourier imaging, 253-255
properties of, 49 Partial saturation, 204-207, 228-232, 231f, 235
spin of, 49-50 Partial saturation pulse sequence, 60, 204f, 208,
spin quantum number of, 41t 229
Number of signal acquisitions, 118-119 Partial saturation spin echo sequence, 208
“Nyquist ghost,” , 303 Partial volume averaging, 396
Nyquist theorem, 107b, 108-109 Parts per million scale, 144, 199b, 264, 311-312
Passive shielding, 136
Patient(s)
O at-risk occupational groups, 439-440
Oblique image, combined gradients for, 150-151, 150f education of, 428
Occipital lobe, 214 evaluation of, 416-419
Occupational exposure to energy fields, 420 auditory concerns, 418-419
Oersted’s experiment, 30-33, 30f-31f biomedical implants, 417-418
Off-resonance artifacts, 401 claustrophobia, 416-417, 416f
Ohm’s law, 24, 24b contrast reactions, 418
180° RF pulse, 48, 52, 61, 78-80, 81f, 232, 248, pacemakers, 417
260-261 preparation of, 426-429, 439-440
Operating console, 116-120 scheduling of, 426
description of, 112, 116-117 Patient couch, 113-114, 114f, 114t
illustration of, 117f Pattern recognition, 216, 217f
508 Index

Perfluorohexybromide, 382-383 Pregnancy, 420, 440

Perfusion, 345 Preventive maintenance, of magnetic resonance imaging
Perfusion imaging system, 430-431
delayed time to peak, 350-353 Primary magnetic coils, 114, 115f
exogenous, 346-347, 346b Projection reconstruction magnetic resonance imaging,
Periodicity, 98 175, 176f
Periventricular lesion, 259f Prostheses, 440
Permanent magnets, 116, 116f, 128-132, 129f-132f, Prosthetic heart valves, 417
132t Proton
Persistent switch, 124 description of, 18
Personnel, 439 magnetic resonant frequencies of, 267f
Petrosal vein, 403f relaxation of, 195f, 196
Phantom images, 433 spinning, 40
Phase coherence, 12, 13f, 71-72 Proton density, 66
Phase contrast magnetic resonance angiography, 340, Proton density images
356, 356f description of, 227
Phase-encoding gradient magnetic field, 177, 200-202, example of, 227f
247-248, 248f, 253-254, 273 Proton density–weighted images
Phase-sensitive reconstruction, 253 definition of, 229
Phase shift effects, 335, 338f description of, 429
Phased array, 321f how to make, 85
Phased array coils, 157-159, 159f Proton spins, 228-229
Phosphorus nuclei, 314-315, 315f Pulsatile flow, 332, 332f
Photoelectric effect, 37 Pulse, 52. See also Radiofrequency pulse
Photon, 35f Pulse amplitude, 119-120
Photopic vision, 215 Pulse sequence
Pixel(s) blipped echo-planar imaging, 299, 300f
description of, 118 cardiac magnetic resonance imaging, 363-365
location of, in slice, 199-200 Carr-Purcell-Meiboom-Gill, 245-246
properties of, 194 definition of, 60
Pixel brightness, 223-224 diagrams
Pixel character, 222-223, 225 description of, 193
Pixel shift, 253b partial saturation, 204-207
Pixel size, 186b diffusion imaging, 357-359
Pixel values, 166 echo-planar imaging, 299f
Planck, Max, 35-36 gradient echo, 273, 273f
Plug flow, 332, 332f gradient echo imaging, 285
Poiled gradient echo imaging, 277 inversion recovery, 62, 62f, 87-88, 89f, 207-208,
Pointillists, 216, 218f 207f, 232-233, 233f
Power of 2, 164-165, 165t magnetic resonance imaging parameters and, 429t
Power of 10, 165t multi-echo spin echo, 61
Power supply multislice spin echo imaging, 210f
gradient coil, 123-124 one-pulse, 60
magnet, 124 partial flip, 60
Precession partial saturation, 60, 204f, 208, 229
definition of, 45 pixel character and, 225
description of, 9-10, 9f precession of steady state imaging fast, 285, 286f
fast imaging with steady state, 282-284, 283f, 288f purpose of, 194-195
frequency of, 49 radiofrequency pulse sequence versus, 246-247
gradient effect on, 200-204 saturation recovery, 60, 60f
MXY, 53, 54f selection of, 118
of nuclear magnetic moment, 42f spin echo, 61, 208-210
steady state free, 364 diagram of, 77f, 209f
Precession of steady state imaging fast, 285, for T1-weighted images, 86
286f-287f stimulated echoes produced by, 63f
Precessional frequency, 335 summary of, 64b
 Index 509

Purcell, Edward, 4 Radiofrequency pulse (Continued)

Pure magnetic resonance images, 226-227 soft, 51-52
XY magnetization, 52-53
Q Radiofrequency pulse sequences
definition of, 60
Quadrature coils, 152-153, 153f-154f
inversion recovery, 62, 62f
Quadrature detection, signal-to-noise ratio affected by,
multi-echo spin echo, 61
153f
one-pulse, 60
Quadrigeminal plate, 222f
partial flip, 60
Quality control, 433-435, 433t, 434f-435f, 435t
partial saturation, 60
Quantization, 166-167
pulse sequence versus, 246-247
Quantum mechanics
saturation recovery, 60, 60f
definition of, 39
spin echo, 61
description of, 40-41
stimulated echoes produced by, 63f
Quantum theory, 36, 36b
summary of, 64b
Quench, 139-142, 142f, 432-433
Radiofrequency radiation, 4
Radiofrequency shimming, 330, 393
R Radiofrequency spoiling, 282
R-factor, 325 Radiograph. See also X-ray(s)
Rabi, Isador, 4 contrast resolution of, 220
Radians per second, 183 definition of, 225
Radiofrequency images associated with, 218-221, 220f
biological effects of, 410 Radiologic technologists
composition of, 47 accreditation of, 433, 434f
uses of, 37 pregnancy in, 420
Radiofrequency amplifier, 123 scheduling of, 425-426
Radiofrequency coils, 34, 34f, 59, 115f staffing of, 425
cardiac magnetic resonance imaging, 362 Rapid acquired relaxation enhanced imaging, 255-259,
description of, 114-115, 143-144 258f
Radiofrequency emission, 4 Read gradient, 202, 299
Radiofrequency fields, 409-410, 414 Read gradient magnetic field, 273-274, 274f
Radiofrequency probe, 151-159 Readout gradient magnetic field, 149f
body coils, 153-154 Received bandwidth, 188
definition of, 151 Receiver gain calibration, 120
design of, 151 Reed switch, 417
head/extremity coils, 154, 155f Reference frames, 46-48
matrix coils, 157-159 rotating, 47-48, 47f-48f
noise detection by, 153 stationary, 46-47, 46f
phased array coils, 157-159, 159f Refocused gradient echo imaging, 282-284
quadrature coils, 152-153, 153f-154f Refocusing pulse, 235
simple, 151, 151f Relativity, theory of, 36b
surface coils, 154-157, 156f-157f Relaxation times, 12-14
Radiofrequency pulse description of, 35
diagrams of, 56, 56f longitudinal, 134
duration of, 52 T1
flip angle of, 292-293 of cerebrospinal fluid, 229
hard, 51-52 contrast agent effects on, 378
intensity of, 52 description of, 12-14, 66-69, 67f-68f, 68t, 74
at Larmor frequency, 51f T2, 69-74
long, 198 cerebrospinal fluid, 235f, 236
in magnetic resonance imaging, 52 contrast agent effects on, 378
90°, 48, 52, 56, 61f, 86, 229, 248 measurement of, 82f
180°, 48, 52, 61, 78-80, 81f, 232, 248, phase coherence, 71-72
260-261 summary of, 74
purity of, 198 T1 relaxation versus, 71
sinc functions, 185 in tissue, 70t
510 Index

Repetition time, 229 Shadowgram, 37

artifacts and, 249 Shielding, 37
cerebrospinal fluid and, 230 Shim coils
definition of, 60, 118, 204-205 description of, 115-116, 115f
free induction decay and, 229 parts per million scale, 144
intermediate, 237f Shimming
length of, 60 description of, 144-146
pixel appearance and, 230, 231t gradient coils for, 145
spin echo pulse sequence, 236, 238f-239f radiofrequency, 330, 393
Representational images, 217-218 Short time inversion recovery sequences, 251
Repulsion, 26-27, 27f Signal attenuation, 357, 357b, 358f
Request form, 426, 427f Signal-to-noise ratio, 187-188
Resistive electromagnet imaging system, 115-116 calculation of, 205b
Resistive electromagnets, 133-134, 133f, 133t definition of, 187
Resonance, 47 description of, 85, 118
Resonant frequency for fat, 264 parallel imaging, 325-327, 326f-327f
Retinal receptors, 214 quadrature detection effects on, 153f
RF. See Radiofrequency of surface coil, 156
Rods, 214-215 white noise effects on, 188
Roentgen, 36 Simple radiofrequency probes, 151, 151f
Roentgen, Wilhelm Conrad, 4 Sin x/x, 93
Room temperature coils, 144-146 Sinc function, 93, 100f, 101, 185
Rotating frame of reference, 47-48, 47f-48f, Sinc pulse, 204
200 Single-phase current, 24
Single photon emission computed tomography, 346
Sinogram, 162-163
S Slew rates, 146, 146b, 147f, 359, 362
Saddle coil, 152f Slice, pixel location within, 199-200
Safety Slice cross-talk, 207
considerations for, 415-419 Slice selection, 196-199
magnetic resonance imaging system, 435-439 Slice selection gradient, 149, 247
patient evaluation, 416-419 Slice thickness, 119
Sampling, 106-109, 108f SNR. See Signal-to-noise ratio
Sampling bandwidth, 265, 267f Sodium, 315, 316f
Sampling process, 168 Soft radiofrequency pulse, 51-52
Sampling theory, 107b Solenoid, 31-32, 32f
SAR, 410 Source equation, Fourier transform of, 94
Saturated spin, 52-53 Source-to-image receptor distance, 6
Saturation recovery, 229 Space–spatial frequency, 146b
Saturation recovery pulse sequence, 60, 60f Spark artifact, 188-189
Scalar quantity, 43 Spatial frequency, 181-185
Scheduling, 425-429 contrast, 188
Scotopic vision, 215 definition of, 182, 201
Secondary magnetic coils, 114-115, 115f high, 291
Segmentation horizontal, 174
definition of, 293 of imaging modalities, 171t
of Fourier lines, 293, 293f impact of, 186-189
k-space, 363 line pairs, 181, 181f
Segmented k-space, 191 line patterns, 170-171, 171f
Self-shielding gradient coils, 148-149 low, 291
Semiconductors, 24, 25t magnitude of, 163
SENSE, 321-322, 324 patterns and order, 190-191, 191f
Sensitivity, 7, 376 projection reconstruction measurements of,
Sensitivity encoding, 320-322 190f
Seperability, 98 signal-to-noise ratio, 187-188
Sequencing system, 124 types of, 171
 Index 511

Spatial frequency domain Spin echo imaging (Continued)

description of, 95, 163, 170-174 rapid acquired relaxation enhanced imaging,
image matrix, 172-173, 172f 255-259, 258f
importance of, 180 stimulated echoes, 261
map of, 174 T2-weighted images, 262f
polar sampling of, 176f gradient echo imaging and, 295
rectangular sampling of, 177f inversion recovery imaging, 250-253
source of, 181 k-space in, 249f, 254
in spin echo imaging, 248 multislice, 364-365
Spatial frequency filtering, 294-295 signal intensity in, 429
Spatial localization, 174-177 spatial frequency domain in, 248
Fourier transform and, 103-105 Spin echo imaging time, 366b
gradient magnetic fields used to create, 175 Spin echo pulse sequence, 61, 208-210, 245-250
Spatial resolution description of, 234-240, 235f-240f
cardiac magnetic resonance imaging, 371 diagram of, 77f, 209f, 247f
description of, 6-7, 6t, 186 inversion recovery, 252f
equation for, 148b, 157b repetition time for, 236, 238f-239f
field of view effects on, 119 for T1-weighted images, 86
measurement of, 170 Spin echo time, 118
pixel size and, 173 Spin ensemble, 67
in radioisotope imaging, 371 Spin-lattice relaxation, 333
of surface coils, 157 Spin-lattice relaxation time. See also T1 relaxation
Specific absorption rate, 118, 329 time
Specificity, 7, 376 definition of, 68
Spectrum, 306 determination of, 207
Spin Spin quantum number
aggregate of, 43 definition of, 40
definition of, 40 of medically important nuclei, 41t
saturated, 52-53 Spin state, 40, 228f
Spin echo, 60-62 Spin warp method, 201
blood flow in, 334 Spinning magnetic field, 10, 10f
definition of, 208 Spiral filling, 368
description of, 184f Spoiled, 278-279
dual echo fast acquisition interleaved, Spoiled gradient echo imaging, 282-284,
255-256 283f
late, 240f Spoiler pulses, 261
multi-echo, 83f, 235 Staffing, 424-425, 424t
T2-weighted images using, 301f Static magnetic fields, 362
weighting of, 237t description of, 407
zero order, 259-260 effects of, 412-413
Spin echo diffusion imaging, 356f health hazards of, 407
Spin echo imaging long-term effects of exposure to, 408
acronyms used in, 245t Stationary frame of reference, 46-47, 46f
echo planar imaging versus, 256 Steady state
fast free induction decay in, 286f
clinical applications of, 261 gradient echo imaging and, 275-278, 280f
concerns in, 259 Steady state free precession, 364
contiguous slice, 255-256 Stern, Otto, 4
contrast enhancement with, 260-261 Stimulated echo, 62-64, 63f, 261, 282-283
definition of, 255-256 Superconducting electromagnets, 134, 134t,
description of, 253-260 137
effective echo time in, 258, 258f Superconducting magnetic resonance imaging system,
gradient echo imaging and, 295 113-115, 113f-114f, 128
J-coupling, 261 Superconductivity, 134-139, 135f-139f
k-space ordering in, 259-260 Superconductor, 24, 25t
partial Fourier imaging, 253-255 Superparamagnetic contrast agents, 377-379
512 Index

Surface coils, 154-157, 156f-157f Three-dimensional Fourier transform

Surgical clips, 440 description of, 16, 175
Susceptibilities, 74 magnetic resonance imaging, 177
Susceptibility artifacts, 83-84, 328 Three-dimensional magnetic resonance angiography,
341-343, 342f-343f, 372-373
Threshold, nonlinear relationship, 406, 406f
T Threshold intensity, 406
T1 relaxation time. See also Spin-lattice relaxation Time-of-flight magnetic resonance angiography, 288,
time 339-340, 372
of cerebrospinal fluid, 229 Tissue
contrast agent effects on, 378 mobile hydrogen nuclei in, 67t
description of, 12-14, 66-69, 67f-68f, 68t, 74 T1 relaxation times in, 68t
T1p relaxation time, 68-69, 69f, 74 T2 relaxation times in, 70t
T2 relaxation time, 69-74 Tissue contrast, 89t
cerebrospinal fluid, 235f, 236 Tissue diffusion, 354-357, 355f
contrast agent effects on, 378 Tissue perfusion, 382
measurement of, 82f Tissue tagging, 364-365
phase coherence, 71-72 Transform
summary of, 74 definition of, 92, 92b
T1 relaxation versus, 71 Fourier. See Fourier transform
in tissue, 70t inverse, 92b, 185, 189f
T2 star (T2*) relaxation uses of, 93-94
definition of, 71 variable, 92b
description of, 72-74, 347 Transformer, 34
T1 weighted images Transient magnetic fields, 408-409, 411-414
description of, 227 Transmitted bandwidth, 188
dynamic range of, 88 Transverse magnetization, 45, 247-248, 292-293,
example of, 227f 300-302
how to make, 85-88, 86f-87f Transverse relaxation, 53-54
magnetic resonance angiography use of, Transverse relaxation time, 12-14, 69
382 Transverse steady state, 284
net magnetization, 85 Triglycerides, 313
spin echo pulse sequences for, 86 Truncation artifact, 396-397, 397f
T2-weighted images versus, 88 Tumor localization and characterization, contrast agents
T2 weighted images for, 381-382
description of, 227 Turbo imaging, 290-295
example of, 228f two-dimensional, 294
fast spin echo, 261, 262f Turbo spin echo, 255-256
how to make, 77-82, 77f-82f TurboFLASH, 291, 291f
signal intensity on, 380 Turbulence, 332-333, 334f
spin echo for, 301f Two-dimensional Fourier transform, 16
T1-weighted images versus, 88 description of, 174, 176-177
T2* weighted images magnetic resonance imaging, 176-177
applications of, 83-84 motion effects on, 178f
how to make, 82-84 Two-dimensional magnetic resonance angiography, 340,
production of, 83 341f
signal intensity on, 380 Two-dimensional turbo imaging, 294
signal losses in, 84f Two-pulse sequence, 232
Temperature scales, 138b, 139t
Temporal bones, 392-393
Temporal frequency, 182-183 U
Temporal resolution, 243, 244f, 245 Unaliasing by Fourier encoding the overlaps in the
Tesla, 9, 29, 31 temporal dimension, 330
Tetramethylsilane, 264, 312 UNFOLD. See Unaliasing by Fourier encoding the
Thallium imaging, 361 overlaps in the temporal dimension
Thermogram, 219f Upfield, 312
 Index 513

V X-ray(s). See also Radiograph

history of, 4
Variable transform, 92b
images produced by, 218-221, 220f
Vector, 9
mechanism of action, 37
definition of, 43
principles of, 4
net magnetization, 51, 53, 53f
production of, 37
Vector diagrams, 43-45, 47
uses of, 37, 37f
Velocity profile, 332, 332f
X-ray attenuation coefficient, 6
Ventricular fibrillation, 413-414
XY magnetization, 52-53
Visible light, 3-4, 36-37
Visual acuity, 215
Y
Visual images, 214, 215f
Volt, 21-22 Y-axis, 132f
Voltage, 21-22 Y gradient coils, 150, 150f
Voltage range, 166
Z
W Z-axis, 9
Weighted images description of, 195
description of, 228-240 net magnetization along, 45, 51
inversion recovery, 232-234, 234f selective excitation along, 197f
partial saturation, 228-232, 231f Z-axis magnetization, 292-294
spin echo, 234-240, 235f-240f Z gradient coils, 149, 149f
T1. See T1 weighted images Z gradient magnetic field, 103-104, 104f
T2. See T2 weighted images Zero frequency artifact, 397, 398f
White noise, 188 Zero order spin echo, 259-260
Windowing, 223, 224f Zeumatography, 5
Workstation consoles, 122 Zipper artifact, 394-395, 397
Zoning, of MRI suite, 437-439, 438f
X
X-axis, gradients along, 196
X gradient coils, 149-150, 149f
This page intentionally left blank
This page intentionally left blank
This page intentionally left blank