DIGITAL IMAGE PROCESSING IN MEDICINE

Doru Cioata,
University of Craiova
Faculty of Automation, Computers and Electronics
5, Lapusului Road, Craiova, Romania
Cristian Iacob,
Carol Davila University of Medicine and Pharmacy
37, Dionisie Lupu Road, Bucharest, Romania
Mihaela Cioata, Roxana Timplaru, Mihaela Papa,
Computer Science High-School “Stefan Odobleja”
13,Electroputere Road, Craiova, Romania

Abstract: This paper presents the use of digital Digital image processing allows the enhancement of
image processing and digital integrated systems in the image’s features of interest while attenuating details
medicine, highlighting the advantages offered. It is that are irrelevant in the given context, and then extract
proposed a complete system that can be implemented or increase the amount of useful information.
in the medical domain. In particular we present a
well known image processing algorithm, unsharp Such a field of interest with a huge amount of
mask. A demonstration of the enhancement applicability for digital image processing is medicine.
capabilities of the unsharp mask algorithm is done The advantages of digital images over analog images are
by applying it on a blurry magnetic resonance obvious. Digital systems are cheaper, more flexible and
obtained image. The result is an image with information can be shared more easily. The information
sharpened edges, clearer details and lower level of can be processed and interpreted in most of the cases
noise. almost fully automated, and so the speed is highly
increased and costs are drastically lowered.

Keywords: Digital Image Processing, MRI (Magnetic Still, complete automation represents a dream.
Resonance Imaging), Unsharp Mask, Low-pass Filter, Human factor remains irreplaceable in information
High-pass filter and Gaussian Blur interpreting, diagnostic settings and, ultimately, in the
future interventions on the patient.

1. INTRODUCTION 2. THE USE OF DIGITAL IMAGING IN

MEDICINE
There has been a renewed interest in digital image
processing in recent years due to a number of factors Digital image processing may be subdivided into
like an increasing availability of powerful three groups as depicted in Figure 1: preprocessing (low
microcomputers and workstations with very large level of processing – preliminary operations used to
capability, a significant improvement in the capabilities filter noise, and to eliminate the unnecessary or
of the image acquisition equipment and display devices unwanted details), data reduction (mid level processing -
and a significant reduction in the cost of computation extraction of useful information, segmentation), feature
and image acquisition. These have led to a rapid analysis (high level processing – ensuring semantic
increase in the application of digital image processing meaning extraction).
techniques such as in astronomy, seismology, medical
applications, radar and so on. The basic task of digital image processing in
medicine is source image enhancement. Images received
Although the mathematical support was well defined, from various medical devices (X-ray, Magnetic
digital image processing was not used due to high Resonance Imaging (MRI), X-Ray based Computer
computing power and large dimension of mass storage Tomography (CT) a.s.o.) - usually not very clear and
devices requested. containing noise - are filtered by using low level image
processing algorithms (denoising, sharpening, edge
The latest evolution in hardware development made detection) with parameter values tweaked for the
these requests affordable enough to make domains that specific problem and enlarging the amount of useful
traditionally used analog imaging to switch to digital information offered to the medical specialist that
images. interprets it. This is fully done by computers.
 The next processing operation is extraction of the to be done) is merging of several data acquisition and
useful information. On the enhanced image there are reconstruction of objects (3D object reconstruction from
applied segmentation and edge linking algorithms for slices). An example of using this technique is esthetical
isolating certain body parts that are subject of interest surgical planning.
(mid level digital image processing algorithms). Simple
analysis of the body parts can be done fully automatic Digital archiving of information offers a large
(e.g. measurement and comparison with the normal number of advantages. Firstly, smaller space is needed
dimensions of a certain body part) and a preliminary for storing data. Secondly, the system allows flexible
diagnosis can be set. This is carried out, depending of and easy access to data (basically from everywhere).
the context, fully automatic or with little human There are also many possibilities to interpret (medical,
intervention. economical, educational, statistical) the stored
information: statistical information obtained easy (e.g.
One of the most complicated issues in digital image comparison of different techniques or different
processing is automatic semantic meaning extraction medication, costs evaluation), comparisons with other
from source images. This implies using of high-level similar cases can be simple, easy observing of the
digital image processing algorithms that are needed to evolution in time of patients with similar problems, the
improve medical diagnostic, operation planning and possibility to monitor the evolution in time of the body
image guided surgery. Human intervention is required in parts that are subject of interest for a patient.
most cases.
Another advantage offered by the digital storing of
Another advantage of digital systems over analog data is easy "sharing" of information. There are many
systems is automatic combination of different categories of “potential users”: medical personnel,
acquisitions or/and different techniques in order to medical research teams, hospital management team,
obtain more relevant information for human analysis university teachers and students and so on.
process.

An already used technique (but much more remains

Source image

Low level algorithms

(preprocessing)
(sharpening, edge detection )

Mid level algorithms

(extraction of useful Digital image
information – segmentation ) processing

High level algorithms

( semantic meaning
extraction )

Classification

Compression

Storage

Figure 1: Digital image processing steps

3. A PROPOSED COMPLETE SYSTEM

Medical Computer
Device 1 1

Database Web Internet

DB Sever
Server

Medical Computer
Device n n

Hospital University
An example of Manager students Surger
potential users (evaluation (didactical (surgical
of the system of costs) planning)
activities)

Figure 2: A proposed integrated system

4. UNSHARP MASK ALGORITHM EXAMPLE

We present here a simple application for use of the classical unsharp masking filter to reduce
enhancement of images received from Magnetic of the noise and sharpen the edges.
resonance imaging (MRI) devices. Magnetic The unsharp mask filter performs a 2D high pass
resonance is a relatively new technique (1990), MRI filter by subtracting a low-pass filtered image from
systems becoming, in clinical diagnosis, a standard the original and adding the result to the original
tool for detecting a variety of tumors, lesions and image. It eliminates homogeneous regions and
abnormalities. The advantage that MRI has over X- highlights edges.
rays and X-Ray based Computed Tomography is that
magnetic resonance signal is able to penetrate radio- Firstly it is obtained a blurred version of the
opaque tissues (e.g. bones) with little attenuation or original image. This is carried out by applying the
distortion. MRI technique offers an excellent soft low-pass filter, in our case Gaussian blur algorithm
tissue contrast and low noise. However, the level of using a small radius. We used a two pixel radius and
the present noise depends on the acquisition time. applied Gaussian blur filter only once. The blurred
This means that real time images will contain a version of the image is then pixel subtracted from the
higher level of noise than non real time MR images. original image and so it is obtained the high pass
component. The output image is obtained by adding
Hence, a basic operation that can be done on MRI the high- pass component to the original image.
is noise removal and sharpening of the boundaries Because the output image could contain also pixels
between regions. MRI images are usually gray level with negative values, it is then normalized. No
(with 256 levels of gray) low resolution (256*256, threshold cutoff was used.
512*512) images. This application demonstrates the
 Blurred image _ Edge image +
Gaussian
Original image (low pass + (high pass + Sharpened image
blur signal) signal)
+ +

Figure 3: The diagram of unsharp mask algorithm processing

Figure 4: Original MRI Figure 5: The image after the unsharp mask
algorithm is applied. Noise is significantly reduced
5. CONCLUSIONS and the boundaries between regions are sharpened.

Medical image processing has become one of the Jain A. K., Fundamentals of Digital Image Processing,
most important fields in medical applications. Prentice Hall, 1989.
Nowadays, basic image processing and visualization
techniques are frequently used. However, there is a great Pratt W. K., Digital Image Processing, 2nd ed. Wiley,
demand for high-level image processing algorithms to 1991.
analyze and visualize anatomical and pathological image
structures in a user-oriented mode, faster and better Russ J. C., The Image Processing Handbook, CRC
compression methods, design and implementation at Press, 1992.
medical insight level of integrated systems and linking
them via Internet to become the ultimate medical tool of Shung K. K., Smith M. B., and Tsui B. M. W.,
modern medicine. Principles of Medical Imaging, San Diego Academic
Press, 1992.
REFERENCES
Sonka M. and Fitzpatrick J., Handbook of Medical
Cho Z. H., Jones J.P. and Singh M., Foundations of Imaging, Vol. II, SPIE Press, 2000.
Medical Imaging, Wiley, 1993.
Sonka M., Hlavac V., and R. Boyle, Image Processing,
Gonzalez R. and Woods R., Digital Image Processing, Analysis, and Machine Vision, 2nd Edition, PWS
Addison-Wesley Publishing Company, 1992 Publishing, 1999.

Haralick R. M. and Shapiro L. G., Computer and Robot Webb S., The Physics of Medical Imaging, IOP
Vision, Addison Wesley, 1992. Publishing, 1992.