Bio-Medical

Image
Analysis
INTRODUCTION
Introduction to Imaging Systems

▪ Of the five senses – sight, hearing, touch, smell and taste – which
humans use to perceive their environment, sight is the most
powerful.
▪ Receiving and analyzing images forms a large part of the routine
cerebral activity of human beings throughout their waking lives.
▪ In fact, more than 99% of the activity of the human brain is
involved in processing images from the visual cortex.

“A picture is worth a thousand words,”

Classification of imaging systems

Fig: A large part of the electromagnetic spectrum showing the energy of one photon, the frequency, wavelength and typical applications of
the different areas of the spectrum.
Credits: Rasmus R. Paulsen, Thomas B. Moeslund, “Introduction to Medical Image Analysis (Undergraduate Topics in Computer Science)”, 1st
Edition, Springer, 2020
Classification of imaging systems
Classification of imaging systems
Optical System Image Acquisition: Sampling and Quantization
Classification of imaging systems
Credits: DIP, Gonzalez
 Computer Aided Detection/Diagnosis

▪ Computer-aided detection (CAD): systems that assist doctors in the

interpretation of medical images.
▪ Imaging techniques in X-ray, MRI, Endoscopy, and ultrasound diagnostics yield
a great deal of information that the radiologist or other medical professional
has to analyze and evaluate comprehensively in a short time.
▪ CAD is an interdisciplinary technology combining elements of artificial
intelligence and computer vision with radiological and pathology image
processing.
▪ Although CAD has been used in clinical environments for over 40 years, CAD
usually does not substitute the doctor or other professional, but rather plays a
supporting role.
 Medical Imaging Systems

▪ Ionizing Imaging: ▪ Non Ionizing Imaging:

▪ Ionizing radiation with sufficient energy to ionize atoms and ▪ Non-ionizing radiation, on the other hand, does not have the
molecules within the body. potential to damage the body directly and the risks associated with
its use are considered to be very low.
▪ Ionizing radiation in medical imaging comprises x-rays and γ-
rays, both of which need to be used prudently to avoid causing ▪ Examples of such radiation are ultrasound, i.e. high-frequency
serious damage to the body and to its genetic material. sound, and radio frequency wave.

▪ Structural medical imaging: ▪ Functional medical imaging:

▪ Structural imaging includes producing images of the anatomy ▪ Functional imaging (or physiological imaging) is a medical
and morphology of the body part generated using computed imaging technique of detecting or measuring changes
tomography (CT), magnetic resonance imaging (MRI), and in metabolism, blood flow, regional chemical composition, and
ultrasound (US) scans. absorption.

▪ Invasive Imaging: ▪ Non Invasive Imaging:

▪ Invasive Imaging performs diagnostic procedures through ▪ Non Invasive Imaging performs diagnostic procedures without
very small incisions or body orifices. any incisions or body orifices.
Non-Invasive and Ionizing
Medical Imaging Systems
X-Ray
▪ X-ray imaging has been used in clinical diagnosis almost from the time of Roentgen’s discovery of x-
rays.

▪ X-rays are a form of electromagnetic radiation, similar to visible light.

▪ Unlike light, however, x-rays have higher energy and can pass through most objects, including the
body.

▪ Medical x-rays are used to generate images of tissues and structures inside the body.

▪ X-rays are generated in an x-ray tube, which consists of an evacuated tube with a cathode and an
anode.
▪ To create a radiograph, a patient is positioned so that the part of the body being imaged
is located between an x-ray source and an x-ray detector.

▪ When the machine is turned on, x-rays travel through the body and are absorbed in
different amounts by different tissues, depending on the radiological density of the
tissues they pass through.

▪ Radiological density is determined by both the density and the atomic number (the
number of protons in an atom’s nucleus) of the material being imaged.

For example, our bones contain calcium, which has a higher atomic number than most
other tissues.

Therapeutic Usage:

Radiation therapy in cancer treatment: X-rays and other types of high-energy radiation can
be used to destroy cancerous tumors and cells by damaging their DNA.

The radiation dose used for treating cancer is much higher than the radiation dose used for
diagnostic imaging.
 CT

▪ Conventional radiographic procedures, those which we have described so far, produce planar images that
are projections of three-dimensional objects onto two-dimensional planes. This results in a considerable
loss of information.

▪ Tomographic imaging, of which x-ray computed tomography (CT) is an example, is a technique that was
developed for producing transverse images, by scanning a slice of tissue from multiple directions using a
narrow fan-shaped beam.

▪ The data from each direction comprise a one-dimensional projection of the object, and a transverse image
can be retrospectively reconstructed from multiple projections.

▪ A computed tomography (CT) scanner looks like a big, square doughnut. The patient is placed within the
aperture of the rotating frame or gantry.

▪ In the third-generation scanners, an x-ray tube mounted on the gantry revolves around the patient, a tightly
collimated fan-beam of x-rays enters the patient, and an arc of detectors on the opposite side, and rotating
synchronously with the x-ray tube, records the intensity of emerging radiation.

▪ The x-ray tube is always on, and readings from the detectors are taken about a thousand times during the
rotation.
▪ The first-generation CT scanner, used for scans of the head,
illustrates many of the principles of the method, and is easier to
visualize (Fig.).

▪ A tightly collimated beam minimizes scatter and radiation dose to

adjacent tissue.

▪ The beam is swept linearly across the patient’s head, and a single
detector, moving synchronously with it, measures the transmitted
radiation at regular intervals; this is known as a scan.

▪ The tube is turned off, the gantry rotated by a small angle (~1°),
and another scan performed. This is repeated to give scans
covering 180°.

▪ The sampling interval and the angle of rotation determine the pixel
size in the reconstructed transverse image, and the collimator
width determines the slice thickness. The motion is known as
translate-rotate.
 Nuclear medicine (NM) imaging:

▪ Nuclear medicine (NM) imaging uses the γ-rays emitted from radioactive isotopes attached to pharmaceutical tracers that
are specific to certain physiological, metabolic and pathological activities, e.g. cerebral perfusion, myocardial perfusion,
cancer.

▪ These radio-labeled pharmaceutical tracers are ingested or injected into the body where they are circulated and/or
metabolized.

▪ The γ-rays which they emit during radioactive decay pass out of the body and are collected by detectors (gamma cameras)
placed around the patient; these measure the distribution of the tracer within the body, and produce images which show the
functional or metabolic activity in the relevant organs.

▪ There are three basic imaging modalities in nuclear medicine.

▪ Projection studies, called planar scintigraphy, are analogous to projection radiography; all depth information is lost. A
single gamma camera, or a dual-head gamma camera to take anterior and posterior images simultaneously, is used to
detect the emitted γ-rays.
▪ Tomographic imaging, called single-photon emission computed tomography (SPECT), uses a rotating gamma camera to
obtain projection images from multiple angles, which are used to reconstruct cross-sectional images.
▪ Positron emission tomography (PET) detects pairs of 511 keV gamma photons, emitted when positrons are annihilated.
Nuclear medicine (NM) imaging:

▪ What are radioactive tracers?

▪ Radioactive tracers are made up of carrier molecules that are bonded tightly
to a radioactive atom.

▪ Some tracers employ molecules that interact with a specific protein or sugar
in the body and can even employ the patient’s own cells.

▪ For example, in cases where doctors need to know the exact source of
intestinal bleeding, they may radiolabel (add radioactive atoms) to a sample
of red blood cells taken from the patient.

▪ They then reinject the blood and use a SPECT scan to follow the path of the
blood in the patient.

▪ Any accumulation of radioactivity in the intestines informs doctors of where

the problem lies.
 Fig: Combined PET/CT (right) of a mouse provides a more
complete view of the spine than CT (left) or PET (middle) alone.

Courtesy: www.nibib.nih.gov.in
Single Photon Emission Computed Tomography
(SPECT)

▪ SPECT imaging instruments provide three-dimensional (tomographic) images of the

distribution of radioactive tracer molecules that have been introduced into the patient’s
body.

▪ The 3D images are computer generated from a large number of projection images of the
body recorded at different angles.

▪ SPECT imagers have gamma camera detectors that can detect the gamma ray emissions
from the tracers that have been injected into the patient.

▪ The cameras are mounted on a rotating gantry that allows the detectors to be moved in a
tight circle around a patient who is lying motionless on a pallet.
Fig: SPECT machine performing a total body bone scan. The patient lies
on a table that slides through the machine, while a pair of gamma
cameras rotate around her.
Positron Emission Tomography (PET)

▪ Positron emission tomography, PET, is the most recent nuclear medicine imaging
technique: in common with the others, it measures physiological function (e.g. perfusion,
metabolism), rather than gross anatomy.

▪ PET scans also use radiopharmaceuticals to create three-dimensional images. The main
difference between SPECT and PET scans is the type of radiotracers used.

▪ While SPECT scans measure gamma rays, the decay of the radiotracers used with PET
scans produce small particles called positrons.

▪ A positron is a particle with roughly the same mass as an electron but oppositely
charged. These react with electrons in the body and when these two particles combine,
they annihilate each other.

▪ This annihilation produces a small amount of energy in the form of two photons that
shoot off in opposite directions.

▪ The detectors in the PET scanner measure these photons and use this information to
create images of internal organs.
Fig: During a positron emission tomography (PET) scan, patient will lie
on a narrow table that slides into a doughnut-shaped hole. The scanner
takes about 30 minutes to produce detailed images of metabolic
activity in your tissues and organs.
Non-Invasive and Non-
Ionizing Medical Imaging
Systems
Ultrasound imaging

▪ Ultrasonic imaging uses high-frequency (~1–10 MHz) sound waves and their echoes to produce images that
can demonstrate organ movement in real time.

▪ Unlike electromagnetic waves, such as x-rays and γ-rays, ultrasound is non-ionizing and, as such, is considered
safe at the intensities used in clinical imaging systems.

▪ Ultrasound images are constructed by calculating the time taken for ultrasound pulses to travel into the body
and return, after reflection off a tissue surface.

▪ Ultrasound pulses, generally 1–5 μs long, are generated from an ultrasound transducer comprising a
piezoelectric crystal, such as lead zirconate titanate (PZT), sandwiched between a pair of electrodes.

▪ A small sinusoidal voltage applied to the crystal causes it to resonate, producing sound waves as its surfaces
move backwards and forwards; the crystal also detects ultrasound waves by producing a varying electrical
signal.

▪ A typical pulse, just over 1 μs in duration with a frequency of 5 MHz, contains 5 cycles of the wave.

▪ Pulses might be separated by perhaps 1 ms, resulting in a pulse repetition rate of 1000 Hz; they travel through
soft tissue at a speed of about 1540 m s−1 compared with the speed of sound in air of about 330 m s−1
 Ultrasound (US)

How Does Ultrasound Work? - YouTube

Ultrasound
▪ Sound waves are longitudinal waves, i.e. the particles of the material move back and forth in the same direction
that the wave is traveling.

▪ The speed of sound in a material, v, is characteristic of that material and depends on the density of the material, ρ,
and its compressibility, К.

▪ The intensity reflection coefficient, R, at a boundary which is smooth compared to the ultrasound wavelength and
perpendicular to the direction of wave propagation, is given by;

▪ where Z1, Z2 are the acoustic impedances of the materials to either side of the surface
 B-mode Ultrasound

▪ B (brightness)-mode ultrasound is more common, and is used to produce a two-dimensional tomographic or

slice image of anatomical structure by sweeping the beam repeatedly back and forth through the patient’s body.

▪ Each sweep is used to form a single vertical line of the B-mode image comprising a series of bright dots.

▪ The brightness of the dots is determined by the intensity of the reflected echoes, after correcting for attenuation
along the path by time gain compensation.
 Magnetic resonance imaging

▪ Magnetic resonance imaging (MRI) is a non-ionizing technique that uses radiofrequency (200MHz–2 GHz)
electromagnetic radiation and large magnetic fields (around 1–3 tesla (T), compared with the Earth’s magnetic field of
about 0.5 × 10–4 T).

▪ The large magnetic fields are produced by superconducting magnets, in which current is passed through coils of
superconducting wire whose electrical resistance is virtually zero.

▪ MRI images provide anatomical and physiological details, i.e. structure and function, with full three-dimensional
capabilities, excellent soft tissue visualization, and high spatial resolution (~1 mm).

▪ Scans last several minutes, rather than a few seconds as in x-ray CT, so that patient motion can be a problem.

▪ Furthermore, MRI scanners are several times as costly as a CT scanner because of the expensive superconducting
magnet required.
 Magnetic resonance imaging

▪ Photo courtesy of Philips Healthcare

Magnetic resonance imaging

▪ The use of nuclear magnetic resonance principles produces extremely detailed pictures of the
body tissue without the need for x-ray exposure and gives diagnostic information of various
organs.

▪ Measured are mobile hydrogen nuclei (protons are the hydrogen atoms of water, the 'H' in
H20), the majority of elements in the body.

▪ The precessional frequency of nuclei of a substance placed in a static magnetic field B is

calculated from the Larmor Equation:

ω = γB

where ω is the Larmor frequency in MHz, γ is the gyromagnetic ratio in MHz/tesla and B is the
strength of the static magnetic field in tesla.
▪ The gyromagnetic ratio (γ) (MHz/T) for a few commonly measured or imaged isotopes are:
▪ H-1 (Hydrogen) 42.58
▪ F-19 (Fluorine) 40.05
▪ Na-11 (Sodium) 11.26
▪ P-31 (Phosphorus) 17.24
Magnetic resonance imaging

▪ The use of nuclear magnetic resonance principles produces extremely detailed pictures of the body
tissue without the need for x-ray exposure and gives diagnostic information of various organs.

▪ Measured are mobile hydrogen nuclei (protons are the hydrogen atoms of water, the 'H' in H20), the
majority of elements in the body.

▪ The precessional frequency of nuclei of a substance placed in a static magnetic field B is calculated
from the Larmor Equation:

ω = γB

where ω is the Larmor frequency in MHz, γ is the gyromagnetic ratio in MHz/tesla and B is the strength
of the static magnetic field in tesla.

▪ Only a small part of them contribute to the measured signal, caused by their different alignment in
the magnetic field.

▪ Protons are capable of absorbing energy if exposed to short radio wave pulses
(electromagnetic energy) at their resonance frequency.
Magnetic resonance imaging

▪ After the absorption of this energy, the nuclei release this energy so that they return to their initial state
of equilibrium.

▪ This transmission of energy by the nuclei as they return to their initial state is what is observed as
the MRI signal.

▪ The subtle differing characteristic of that signal from different tissues combined with complex mathematical
formulas analyzed on modern computers is what enables MRI imaging to distinguish between various organs.

▪ Any imaging plane, or slice, can be projected, and then stored or printed.

▪ The measured signal intensity depends jointly on the spin density and the relaxation times (T1 time and T2
time), with their relative importance depending on the particular imaging technique and choice of interpulse
times.

▪ Any motion such as blood flow, respiration, etc. also affects the image brightness.
Magnetic resonance imaging

T1-weighted T2-weighted
▪ T1-weighted images are short TR (repetition time) (300– ▪ T2-weighted images are long TR (1800–2500 ms) and
1000 ms) and short TE (echo time) (10–30 ms) and long TE (40–90 ms), sensitive for detecting fluid and
edema.
provide excellent anatomic detail.
▪ The timing of radiofrequency pulse sequences used to
▪ The timing of radiofrequency pulse sequences used to make T2 images results in images which highlight
make T1 images results in images which highlight fat fat AND water within the body.
tissue within the body.
▪ Conversely, T2-weighted images are produced by using
longer TE and TR times.
▪ The contrast and brightness of the image are
predominately determined by T1 properties of tissue.
▪ T2 images – 2 tissue types are bright – FAT and WATER

▪ T1 images – 1 tissue type is bright – FAT

Optical Imaging:

▪ Optical imaging uses light and special properties of photons to obtain detailed images of
organs, tissues, cells and even molecules. The techniques offer minimally or non-
invasive methods for looking inside the body.

▪ Optical imaging significantly reduces patient exposure to harmful radiation by using non-
ionizing radiation, which includes visible, ultraviolet, and infrared light.

▪ Endoscopy

▪ Optical Coherence Tomography

▪ Diffuse Optical Tomography (DOT)

▪ Photoacoustic Imaging
 Eg: Celiac Disease

▪ Malabsorption
▪ Inherited
▪ Autoimmune disease-> Gluten intolerance
▪ 1% Prevalence in Population
 SYMPTOMS OF THE DISEASE

UNEXPECTED WEIGHT LOSS Nausea and Vomiting Protein and Vitamin

Malabsorption

Weakness STOMACH & Abdominal Pain CHRONIC DIARRHEA

Endoscopic Imaging
Introduction

▪ The human eye is one of the vital and complex organs which provides the ability to see and
perceive the surrounding world.

▪ Due to the aging population and increase in the prevalence of diabetes, diseases like age-
related macular degeneration (AMD) and Diabetic Macular Edema (DME) became reasons
for the majority of vision loss [1, 2].

Fig. 1. Anatomy of human eye

 Retinal Disorders

▪ Macular degeneration - a disease that

destroys your sharp, central vision

▪ Diabetic macular edema.

▪ Retinal detachment - a medical

emergency, when the retina is pulled away
from the back of the eye.

▪ Symptoms:

-Blurry or double vision Rings

-Flashing lights, or blank spots Figure 2: Graphical representation of the cross section view of
the macular edema.
-Pain or pressure in one or both of your eyes
Cystoid Macular Edema

▪ The World Health Organization’s recent report1 on effective coverage of eye care states
that:
▪ Currently, an estimated 94 million people aged 50 years and over have moderate-to-
severe distance vision impairment or blindness, while at least 826 million people have
distance- or near-vision impairment.

▪ Cystoid macular edema (CME) is a painless disorder which affects the central retina or
macula [1].

▪ Retinal cysts result from pathological accumulation of fluid in the retina.

▪ Cysts appear as discrete cavities in retina.

1Report of the 2030 targets on effective coverage of eye care, WHO, 12 October 2022, ISBN: 9789240058002
https://www.who.int/publications/i/item/9789240058002
 Cystoid Macular Edema

Figure 3. Fluid leakage affecting the retina as a result of neovascularization in age-related macular degeneration (AMD).

Image Courtesy: The Angiogenesis Foundation - One Broadway, Cambridge, MA 02142, [email protected]
Cystoid Macular Edema
 Diagnosis of Cystoid Macular Edema

▪ Cystiod Macular Edema can be diagnosed using visual acuity test and Amsler grid test.

▪ Slit Lamp Test using a 78 or 90 diopter aspheric lens can be used for visualizing thicken
areas or cystic fluids.

▪ Imaging techniques can also be use for detection of cystic fluids in retina:

• Color Fundus Photography (CFP).

• Fundus Fluorescein Angiography (FA).

• Optical Coherence Tomography (OCT).

Cross Section View of Retina

Fundus Photography of Enface view of Retina

Fundus Photography of Enface view of Retina Florescence Angiography of Enface view of Retina
 Optical Coherence Tomography (OCT)
▪ Optical Coherence Tomography (OCT) is a leading non-invasive imaging technique utilized to
acquire cross-sectional retinal imaging in ophthalmology.

▪ However the large annotated dataset required to achieve better image analysis.

Fig. OCT Image Acquisition and Sample B-Scan.

Optical Coherence Tomography (OCT)
Optical Coherence Tomography (OCT)
Optical Coherence Tomography (OCT) Scan
 Tethered capsule endomicroscopy

Gora, M. J., Quénéhervé, L., Carruth, R. W., Lu, W., Rosenberg, M., Sauk, J. S., … Tearney, G. J. (2018). Tethered
capsule endomicroscopy for microscopic imaging of the esophagus, stomach, and duodenum without sedation
in humans (with video). Gastrointestinal Endoscopy, 88(5), 830-840.e3.
https://doi.org/10.1016/j.gie.2018.07.009
 Gut Probe in a Pill

▪ 2019 | MIT Technology Review

TCE OCT Imaging
G. N. Girish, Leon Alarcon, Paola A., Barrios,
Amilcar, David O. Otuya, Jing Dong, Sarah L.
Giddings, Nitasha G. Mudalaje, Catriona N. Grant,
Joseph A. Gardecki, Norman Nishioka, Christopher
Damman and Guillermo J. Tearney, “Non-invasive
tethered capsule endomicroscopy imaging
biomarkers for the diagnosis of the celiac disease
from unsedated patients", in Endoscopic
Microscopy XVI. Vol. 11620. International Society
for Optics and Photonics, SPIE. . Mar 2020.
Image Types

▪ Gray Scale Images:

To transform the information from the sensor into an image, each cell content is
now converted into a pixel value in the range: [0, 255].
Such a value is interpreted as the amount of light hitting a cell during the
exposure time.
This is denoted the intensity of a pixel. It is visualized as a shade of gray denoted a
gray-scale value or gray-level value ranging from black (0) to white (255)

▪ Binary Images:
In a binary image, there are only two possible pixel values. Unless otherwise noted
we shall talk about 0-pixels and 1-pixels.
Often the set of 0 value pixels are interpreted as the background and the 1 value
pixels as the foreground
Image Types

▪ Label Images:
Label images are typically the result of image segmentation, BLOB analysis, or pixel
classification algorithms.
A typical example is that the pixel value tells which object the pixel belongs to.
If an image contains a man and a dog, the pixel values could be 0 for pixels belonging to the
background, 1 for pixels that constitute the man, and 2 for dog pixels.
▪ Multi-spectral Images:
In a gray-level image there is one measurement (the intensity) per pixel and in color images
there are three values per pixel (the red, green, and blue.
Images containing information captured at multiple spectral wavelengths are called multi-
spectral images or multichannel images.
Images captured by satellites are often multi-spectral.
Furthermore, there is extensive research in multi-spectral cameras for diagnostic purposes as, for
example, early skin cancer diagnostics
Image Types

▪ 16-Bit Images:
An example is modern medical scanners, where the pixel value does
not represent the intensity of incoming light, but an underlying
physical measurement.
In Computed Tomography (CT) scanners, the pixel value is typically
measured in Hounsfield Units (HU) that describe the attenuation of
X-rays in the tissue.
The Hounsfield unit is defined so a pixel that describes air has a
value of 1000 and a pixel describing water has a value of 0.
A typical bone pixel will have a value of 400.
 Image Storage and Compression
▪ Lossless compression
Compression means to represent the image data with less bytes
than the original data. A simple compression method is Run-Length
Encoding (RLE).
Here a sequence of values that have the same value (called a run) is
stored as the count (the number of the values) and the value.

3, 23, 2, 55, 2, 89, 3, 55, 1, 158, 5, 34,

when it is run-length encoded.

The storage is therefore reduced from 16 bytes to 12 bytes.

▪ Run-length encoding is a lossless compression since the original
image data can be computed from the encoded data without any
loss.
▪ A popular image format that stores images using lossless
compression is the Portable Network Graphics (PNG) format.
▪ Images that should be analyzed should always be stored using a
lossless image format.
Lossy Compression and Image Formats

▪ A very popular image format Joint Photographic Experts Group

(JPEG) is based on computing the frequencies in the image and
then only storing the frequencies that the human eye can clearly
see.
▪ It is normally possible to control the amount of data (frequencies)
that should be removed when the image is stored. Compression
and Image Formats.
▪ If the compression ratio is too high, it is possible to see
compression artifacts
DICOM

▪ DICOM is an acronym for Digital Imaging and Communications in

Medicine (DICOM).
▪ As indicated by the title, it is not only a file format but also a standard for
transmitting images over a network.
▪ It is a very complex format that can handle images from almost all
medical imaging equipment.
▪ The header of the DICOM image typically contains patient information
and sometimes also diagnostic information.
▪ Compared to the more standard image formats that store one pixel as
one byte (or three bytes for RGB) DICOM also supports other pixel value
types.
▪ Pixel values are often stored as 16-bit (2 bytes).
Picture archiving and communication systems
(PACS)

▪ A picture archiving and communication system (PACS) is

essentially a network system that allows digital or digitized images
from any modality to be retrieved, viewed and analyzed by a
relevant expert, or by an appropriate expert system, at different
workstations.
▪ DICOM format allows images, and cine-loop images, with
associated patient information and reports, including voice notes,
to be stored and exchanged readily over the network.
▪ PACS systems can be integrated into radiological/hospital
information systems (RIS/HIS), with the inclusion of administrative
information such as billing and inventory
▪ The advantages of PACS systems include:
(i) film-less radiology (no darkroom required, no chemical developers
to purchase, no bulky storage rooms);
(ii) easy access to images, including those from remote sites; (iii)
easy image processing/enhancement;
(iv) easy registration of images from different modalities;
(v) compression of images for quicker communication.

Major disadvantages are the large capital cost of setting it up and

training personnel, and the inevitable difficulties of phasing it in.
Clinical Decision Support System (CDSS)

▪ An overview of clinical decision support systems: benefits, risks,

and strategies for success | npj Digital Medicine (nature.com)