MEDICAL IMAGE PROCESSING

GLOTSOS DIMITRIS
[email protected]

Medical Image and Signal Processing (MEDISP) Laboratory

Department of Biomedical Engineering
Technological Educational Institute of Athens
Greece

May 2017 1
BASIC PRINCIPLES AND DEFINITIONS
About the presentation

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
2
WHY DO WE NEED IMAGE PROCESSING
Image Perception

Source:http://www.qsl.net/w5www/9people.html
3
WHY DO WE NEED IMAGE PROCESSING
Image Perception

4
BASIC PRINCIPLES AND DEFINITIONS
What is a medical image

Definition: -characteristics
Monitoring/recording of the geometric •Anatomical: Static distribution of a certain
physical property, Skeleton.
distribution of certain physical property
•Physiological/Functional: Functionality or
Metabolism of organs, Glucose
consumption in brain.

physical property: physical property:

X-rays attenuation Distribution drug labeled physical property: physical property:
Ultrasound reflection 5
with radioisotope Electromagnetic signals
BASIC PRINCIPLES AND DEFINITIONS
How the image is perceived by the visual
system
-receptors
•Cones: Color detection (6-7 million)
•Rods: Photon detection (90-120 million),
more sensitive than cones, peripheral
vision, night vision

•Digital detectors!!!
0 or 1

Source: http://www.biologymad.com/nervoussystem/eyenotes.htm

6
BASIC PRINCIPLES AND DEFINITIONS
Visual Perception – A low contrast system

-BRIGHTNESS PERCEPTION
•The visual system may differentiate about
64 different grades of brightness – Low
contrast

7
BASIC PRINCIPLES AND DEFINITIONS
What is medical image processing

Definition: Visualize information

Digital image processing refers to
the reversible modification of the
image in the form of a matrix of Reduce
Improve
numerical values. visibility of
[Gonzalez RC, Woods RE, “Digital image processing”,
Prentice Hall; 2002]

Aim:
Enhance, restore, extract,
understand and code
information Adjust and optimize

8
BASIC PRINCIPLES AND DEFINITIONS
Basic image processing steps

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
9
 IMAGE VISUALIZATION
Goal: Transform numbers to •Organize information in a way that the
image observer may understand it: i.e. visualize
the information on the screen

10
 IMAGE VISUALIZATION
Goal: Transform numbers to Methods:
image • Linear Display
• Windows
Applications: • Non-Linear Display
•All digital imaging systems! • Tomographic reconstruction
• Volume rendering

X-ray tube

ADC
PATIENT IMAGE MATRIX
VISUAL DISPLAY
UNIT

DETECTORS

Video tutorial:
https://www.youtube.com/watch?v=9SUHgtREWQc
11
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Generations + History
Year Researcher Milestone

1895 Röntgen x-Ray detection

Source:
1896 First radiographs http://en.wikipedi
a.org/wiki/X-ray

1902 Codman First report that the x-rays

may be responsible for cancer

1913 Coolidge Coolidge x-ray tube

1917 Radon Proof of the tomographic

reconstruction principle

Source:
http://www.crtsite.co
m/page5-2.html

Source:
http://people.csail.mit.edu/bkph/courses/papers/E
xact_Conebeam/Radon_Deutsch_1917.pdf

12
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Generations + History
Year Researcher Milestone

1937 Watson First attempt for transverse

tomography on film
1940 Frank Presentation of the backprojection
method
1961 Oldendorf First model of computed
tomography upon which Hounsfield
Source:
and Cormack were based http://www.impactscan.or
g/CThistory.htm

1963 Cormack Cormack publishes his results from

his custom experimental device for
the reconstruction of a slice of an
object Source:
http://www.smh.com.au/articles/20
04/09/02/1093939066254.html
1968 Hounsfield Hounsfield experimental platform
presentation
1971 Hounsfield First application of CT for brain scan
in hospital (UK)

Source:
http://www.impactscan.or
g/CThistory.htm

13
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Generations + History
Year Researcher Milestone

1973 First application in US

1974 Ledley First body scan

1974 3rd generation systems Source:

http://mobile.nytimes.com/
images/10000000168352
1/2012/07/26/us/robert-s-
1975 2nd and 3rd generation systems ledley-inventor-of-whole-
become commercial body-ct-scanner-dies-at-
86.html?_r=0

1976 5-sec scan

1979 Cormack, Hounsfield Nobel price for Hounsfield and

Source:
Cormack http://tech.snmjournals.org/content/
35/3/115/F10.expansion.html
1981 3-sec scan

1983 800 systems are functional in

US
1985 1-sec scan

Source:
http://www.nobelprize.org/nobel_
prizes/medicine/laureates/1979/
14
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Generations + History
Year Researcher Milestone

1989 Kalender First spiral system

1998 Multiple slice system (4 slices) Source:

http://tech.snmjournals.org/cont
ent/35/3/115/F11.large.jpg

1999 Townsend, Nutt First PET-CT

2002 Multiple slice system (8 and 16 slices)

2007 72.000.000 scans only in US

Source:
http://tech.snmjournals.org/cont
ent/36/2/57/F1.large.jpg

Source: http://www.mdanderson.org/education-and-
research/departments-programs-and-labs/labs/pet-
development-laboratory/research/micro-pet-ct-
camera/index.html

15
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Hounsfield lecture

16
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Hounsfield lecture

Source:
http://en.wikipedia.org/wik
i/Godfrey_Hounsfield

17
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Hounsfield lecture

18
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Hounsfield lecture

19
 IMAGE VISUALIZATION
(HISTORICAL SURVEY)
Hounsfield lecture

20
 IMAGE VISUALIZATION
(INTENSITY WINDOWING)
Simple Window

IVDU REGION 1 : If I x <WS then I y =y1

y1 WE  WS y1  x1
REGION 2 : If WS<=Ix <=WE then 
I x  WS I y  y1
REGION 3 : If I x >WΕ then I y =y2

Iy
WW (Window Width)
WL (Window Level)
WS (Window Start)
WΕ (Window End)
WW
WS  WL 
y2 2
x1 WS Ix WE x2 WW
ΙORIGINAL WE  WL 
2

21
 IMAGE VISUALIZATION
(INTENSITY WINDOWING)
Simple Window

Image matrix (5 bit) Transformation Image (3 bit)

A1= [ 30 29 12 IVDU
A2= [ 7 7 1
I x  30, I y  7
y1
17 12 25 5 1 7
12 8 17] 1 0 5]
I x  29, I y  7
Iy

Numbers (7  0)(12  10) Pixels

I x  12, I y   0  1, 4  1
20  10
(7  0)(17  10)
I x  17, I y   0  4,9  5
y2 20  10
x1 WS Ix WE x2
ΙORIGINAL
I x  25, I y  7

I x  8, I y  0
WW=10, WL=15
WS=10, WE=20

22
 IMAGE VISUALIZATION
INTENSITY WINDOWING - PRACTICAL APPLICATION: CT

Linear display Simple Window

23
 IMAGE VISUALIZATION
INTENSITY WINDOWING - PRACTICAL APPLICATION: CT

Windowed image for lung tissue

Source: LM Fayad et al., Chest CT Window Settings Windowed image for mediastinal tissue
With Multiscale Adaptive Histogram Equalization: Pilot
Study, Radiology 223 (3), 845-852. 6 2002
24
BASIC PRINCIPLES AND DEFINITIONS
Basic image processing steps

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
25
 IMAGE ENHANCEMENT AND
RESTORATION

Goal: Improve/enhance image

quality

Reduce
Improve
visibility of 1. Equalization
2. Filtering
3. Deconvolution

Adjust and optimize

26
 IMAGE ENHANCEMENT AND
RESTORATION
Image content

•Low frequency •High frequency

•Basic shape-overview (no •Details + noise

noise)
•If interested in
•If interested in retaining the enhancing details, then
basic information of the keep high frequencies,
image, then keep low reduce low frequencies,
frequencies, reduce high thus, increase noise—
frequencies, thus, reduce optical effect: edge
noise—optical effect: enhancement 27
blurring
 IMAGE ENHANCEMENT AND
RESTORATION
1. Histogram Equalization: Increase Contrast

Goal: Redistribute image tones

Original Equalized

28
 IMAGE ENHANCEMENT AND
RESTORATION
1. Histogram Equalization: Increase Contrast
Frequency of appearance of image tones
7 4 5 6
1 4 3 5
5 3 2 1
4 3 3 5

29
 IMAGE ENHANCEMENT AND
RESTORATION
1. Histogram Equalization: Increase Contrast

Original Image Transformation Enhanced Image

7 4 5 6 7 3 5 7
1 4 3 5 i h(i) q(i) 0 4 1 5
5 3 2 1 6 2 1 0
0 0 2
4 3 3 5 4 2 3 6
1 2 2

2 1 2

3 4 2

Histogram modification 4 3 2
with histogram 5 4 2
equalization:
6 1 2
redistribution of image
tones in such a way that 7 1 2
each tone will appear
with the same frequency
at the final image

30
 IMAGE ENHANCEMENT AND
RESTORATION
1. Histogram Equalization: Increase Contrast

Original Image Transformation Enhanced Image

7 4 3 5
5 7 4 6 7 3 0 4
4 4 0 6 i h(i) CDFh(i) q(i) CDFq(i) 4 7 3 6
6 3 4 4 0 1 1(0) 2 2 Compression 3 3 0 6
1 0 1(0) 2 4 6 0 3 3
2 0 1(0) 2 6
3 2 3(0) 2 8
4 6 9(3) 2 10 Expansion
5 2 11(4) 2 12
6 3 14(6) 2 14
Histogram modification 7 2 16(7) 2 16 
based on the Cumulative
Distribution Function
(CDF). Fast and
effective method for
increasing image
constrast

31
 IMAGE ENHANCEMENT AND
RESTORATION
HISTOGRAM EQUALIZATION-PRACTICAL APPLICATION: CT
CONTRAST ENHANCEMENT

Source: Wu Z. Q. et al, "The Removal of Blocking Effects Caused by Partially Overlapped Sub-Block Contrast Enhancement", 32
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.103.7458.
 IMAGE ENHANCEMENT AND
RESTORATION
HISTOGRAM EQUALIZATION-PRACTICAL APPLICATION: MAMMOGRAPHY
CONTRAST ENHANCEMENT

Original image Processed with CLAHE

Source: P. Bougioukos et al, "Fast Enhancement of Mammographic Images using Wavelets and CLAHE on a Distributed 33
Processing System", 3rd European Medical and Biological Engineering Conference, Prague, Czech Repurlic, 2005.
 IMAGE ENHANCEMENT AND
RESTORATION
HISTOGRAM EQUALIZATION-PRACTICAL APPLICATION: MAMMOGRAPHY
MICROCALCIFICATIONS ENHANCEMENT

Original image Processed with wavelets

Source: P. Bougioukos et al, "Fast Enhancement of Mammographic Images using Wavelets and CLAHE on a Distributed 34
Processing System", 3rd European Medical and Biological Engineering Conference, Prague, Czech Repurlic, 2005.
 IMAGE ENHANCEMENT AND
RESTORATION
2. Filtering: Reduce noise, enhance texture

Goal: Increase and/or decrease differences

in values between neighbor regions

Original sharpening Filtered

Mask

X -1 -1 -1
-1 9 -1
-1 -1 -1

35
 IMAGE ENHANCEMENT AND
RESTORATION
2. Filtering: Reduce noise, enhance texture

Original matrix Transformation Enhanced Image

30 31 12 9 30 31 12 9
17 12 25 10 17 18 15 10
12 8 17 9 12 18 16 9
31 12 26 22 31 12 26 22

36
 IMAGE ENHANCEMENT AND
RESTORATION
2. Filtering: Reduce noise, enhance texture

Original matrix Transformation Enhanced Image

30 31 12 9 30 31 12 9
17 12 25 10 17 18 15 10
12 8 17 9 12 18 16 9
31 12 26 22 31 12 26 22

37
 IMAGE ENHANCEMENT AND
RESTORATION
FILTERING- PRACTICAL APPLICATION: RADIOGRAPHY
EDGE CRISPERING – HIGH PASS

Original image Sharpened image

Source: A.C. Offiah, et al, Diagnostic accuracy of fracture detection in suspected non-accidental injury: the effect of 38
edge enhancement and digital display on observer performance, Clinical Radiology, 61, 2, 2006, 163-173
 IMAGE ENHANCEMENT AND
RESTORATION
FILTERING- PRACTICAL APPLICATION: RADIOGRAPHY
EDGE CRISPERING – HIGH PASS

Source: Lijun Yin, Scalable edge enhancement with automatic optimization for digital radiographic images, Pattern
Recognition, Volume 37, Issue 7, July 2004, Pages 1407-1422

39
 IMAGE ENHANCEMENT AND
RESTORATION
FILTERING - PRACTICAL APPLICATION: MRI
DENOISING – LOW PASS

Source: Ertürk MA, Bottomley PA, El-Sharkawy A-MM. Denoising MRI Using Spectral Subtraction.
IEEE transactions on bio-medical engineering. 2013;60(6):1556-1562. doi:10.1109/TBME.2013. 40
 IMAGE ENHANCEMENT AND
RESTORATION
FILTERING - PRACTICAL APPLICATION: ULTRASONOGRAPHY

DENOISING – LOW PASS

Original image Denoised image

Source: https://www.lpi.tel.uva.es/node/510 41
 IMAGE ENHANCEMENT AND
RESTORATION
3. Deconvolution: Restore image quality

•Improve visual presentation of the

Goal: Model sources of image degradation
image + prepare image for segmentation
and apply an inverse process to remove
them •Why the image has poor quality?
Why we need to enhance the image?

Degradation model

42
 IMAGE ENHANCEMENT AND
RESTORATION
DECONVOLUTION- PRACTICAL APPLICATION: ULTRASONOGRAPHY
DESPECKLING

Original image Denoised image

Source: Gjenna Stippel, et al, A tissue-specific adaptive texture filter for medical ultrasound images,
Ultrasound in Medicine & Biology, Volume 31, Issue 9, September 2005, Pages 1211-1223
43
 IMAGE ENHANCEMENT AND
RESTORATION
DECONVOLUTION- PRACTICAL APPLICATION: MICROSCOPY

Source: https://www.sccp.sc.edu/Microscopy_Flow_Cytometry_Core_Facilities 44
BASIC PRINCIPLES AND DEFINITIONS
Basic image processing steps

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
45
 IMAGE REGISTRATION AND FUSION
ANATOMY/MORPHOLOGY PHYSIOLOGY/FUNCTIONALITY
-x-ray, CT, MRI, US, Portal Imaging, -SPECT, PET, fMRI, EEG, MEG, special
endoscopy, special protocols (MRA, DSA, protocols (pMRI, fCT, EIT, MRE)
CTA, Doppler)

Goal: Applications:
•integration of complementary information from •Diagnosis (breast and colon cancer),
different imaging modalities
cardiac studies, bone injuries,
•Registration: Bring the modalities inflammatory diseases, neurological
involved into spatial alignment disorders, radiotherapy, surgery
•Fusion: after registration, integrate the
two images into one display

Methods:
• Control point-based
• Moment-based
• Edge-based
• Optimization of similarity 46
 IMAGE REGISTRATION AND FUSION
(MULTIMODALITY IMAGING)
PET/CT
-PET: monitor metabolic activities
-PET has low spatial resolution
-Spatial resolution is enhanced
using CT
Prototype system:- CTI PET
Systems in Knoxville, TN
(today=Siemens Molecular
Imaging)
-Pittsburgh, 1998
-spiral CT scanner (Somatom
AR.SP; Siemens Medical
Source:
Solutions, Forchheim, Germany)
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2777694/pdf/n
ihms67781.pdf + ECAT ART scanner (CTI PET
Systems, Knoxville, TN) 47
 IMAGE REGISTRATION AND FUSION
(MULTIMODALITY IMAGING)
PET/CT
-manufacturers: -applications:
•GE Healthcare •Oncology
(http://www3.gehealthcare.com/en/pro
• PET/CT improves sensitivity in staging
ducts/categories/pet-ct)
(84%) against 63-64% when using only PET
•Philips Medical Systems or only CT (Antoch et al. (2004))
(http://www.healthcare.philips.com/m
• PET/CT presented 50% additional
ain/products/nuclearmedicine/product
information than using only PET or only CT
s/pet/)
(BarShalom et al. (2003)), changing, in this
•Toshiba Medical Corporation way, treatment planning in 14% of patients
(http://medical.toshiba.com/news/pres
• PET/CT contributes in radiotherapy
s-releases/2014/06/09/1940/)
treatment planning ((Herrmann
•Siemens Medical Solutions 2005;Scarfone et al. 2004))
(http://usa.healthcare.siemens.com/mo
lecular-imaging/pet-ct)

48
 IMAGE REGISTRATION AND FUSION
(MULTIMODALITY IMAGING)
PET/MRI

-In some cases, CT doesn’t offer

adequate spatial resolution in soft
tissues
-CT burdens patients with
significant dose
-PET/CT is very good in bone and
lung examinations
-In other cases (i.e. brain,
abdomen) PET/MRI might
improve spatial resolution
-MRI uses non ionizing radiation

Source: http://link.springer.com/chapter/10.1007%2F978-3-
540-72718-7_6#page-1
49
 IMAGE REGISTRATION AND FUSION
(MULTIMODALITY IMAGING)
PET/MRI
-manufacturers: -applications:
•GE Healthcare •Oncology
(http://www3.gehealthcare.com/en/pro
• Pancreas, intestine, uterus, lymphoma
ducts/categories/molecular_imaging/p
et_ct_and_mr_trimodality_imaging) •Cardiology
•Philips Medical Systems • Angiography, myocardium studies
(http://www.healthcare.philips.com/m
•Neurology
ain/products/nuclearmedicine/product
s/ingenuity_tf_pet_mr/) • Epilepsy, neurodegeneration
•Siemens Medical Solutions
(http://www.healthcare.siemens.com/
magnetic-resonance-imaging/mr-pet- Video tutorial:
scanner) https://www.youtube.com/watch?v=3B752nZi1qg

50
 IMAGE REGISTRATION AND FUSION
(APPLICATIONS)
PET/CT
Source:
https://www.aapm.org/me
etings/08SS/documents/
Townsend.pdf

Video tutorial:
https://www.youtube.com
/watch?v=qCT3KQitrCQ

51
IMAGE REGISTRATION AND FUSION
(APPLICATIONS)
PRACTICAL APPLICATION: MRI - SPECT

Brain cancer

Source: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3101722/

52
IMAGE REGISTRATION AND FUSION
(APPLICATIONS)
PRACTICAL APPLICATION: PET - MRI

Source:
https://imaging.stonybrookmedici
ne.edu/imaging-modalities/pet-
mri

Video Tutorial 1:
https://www.youtube.com/watch?
v=h8IwnDYgs0g

Video Tutorial 2:
https://www.youtube.com/watch?
v=d6GPmBMuth4

53
IMAGE REGISTRATION AND FUSION
(APPLICATIONS)
PRACTICAL APPLICATION: PET – CT/MRI
Rectal cancer: Left-sided, FDG-avid
rectal cancer with rectal wall thickening.
Note the improved detectability of the
surrounding infiltration and detectability
of the small locoregional lymph node.

Source: http://www.auntminnieeurope.com/index.aspx?sec=ser&sub=def&pag=dis&ItemID=607257
54
 IMAGE REGISTRATION AND FUSION
(APPLICATIONS)
PRACTICAL APPLICATION: PET – CT/MRI
Abdominal: PET-MRI of
patient with cecal
inflammation. Fused imaging
demonstrates small area of
active inflammatory changes in
the cecum that would be
difficult to diagnose on either
FDG-PET or MR imaging
independently.

Source: http://www.intechopen.com/books/inflammatory-bowel-disease/the-imaging-of-inflammatory-
bowel-disease-current-concepts-and-future-directions
Εαρινό 2014-2015 MTMBIT 2.2 55
BASIC PRINCIPLES AND DEFINITIONS
Basic image processing steps

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
56
 IMAGE SEGMENTATION

Goal: Automated or semi-automated Applications:

delineation of anatomical structures •quantification of tissue volumes,
and other regions of interest
•diagnosis,
•localization of pathology, study of
Methods: anatomical structure,
• Thresholding
•treatment planning,
• Region Growing
•partial volume correction of
• Markov Random fields functional imaging data,
• Deformable models
•computer integrated surgery

57
 IMAGE SEGMENTATION
Thresholding example

Original Image Transformation Segmented Image

100 128 110 90
128 210 220 240 0 0 0 0
0 1 1 1
90 214 210 180 0 1 1 1
90 200 64 30 0 1 0 0

• Fast and effective

• Define a value (the
threshold)
• All pixels with value
greater (or lower) than
the threshold are
marked with 1 (region
of interest), whereas
all remaining values
are marked with 0
(background regions) 58
 IMAGE SEGMENTATION
(APPLICATIONS)
PRACTICAL APPLICATION: MRI

Source: Vaidyanathan, M. Monitoring brain tumor response to therapy using MRI segmentation 59
(1997) Magnetic Resonance Imaging, 15 (3), pp. 323-334.
IMAGE SEGMENTATION
(APPLICATIONS)
PRACTICAL APPLICATION: ULTRASOUND

Source: Gjenna Stippel, et al, A tissue-specific adaptive texture filter for

medical ultrasound images, Ultrasound in Medicine & Biology, Volume 31,
Issue 9, September 2005, Pages 1211-1223
60
 IMAGE SEGMENTATION
(APPLICATIONS)

Source: Source:
Source:
http://www.obgyn.net/articl https://sbarnes.wordpress.
http://sirkan.iit.bme.hu/~cseb/P
es/fetal-face-and-hands- com/reseach-interests/mri-
ublications/WSCG00/
3d-surface-rendering-1 interests/

61
BASIC PRINCIPLES AND DEFINITIONS
Basic image processing steps

Image visualization: Visualize information

Display image on the screen

Image enhancement and restoration: Reduce

Improve
Enhancement of contrast and details,
noise suppression visibility of

Image registration and fusion:

Spatial alignment, multi-modality
image registration and fusion

Image segmentation:
Region extraction
Adjust and optimize
Image classification/recognition:
Computer-aided diagnosis
62
IMAGE CLASSIFICATION/RECOGNITION

Original image Goal: Applications:

•Automated or semi-automated classification 1. Diagnosis
of image content into meaningful recognitions 2. Prognosis
3. Treatment planning
Methods:
1.Pattern recognition
2.Artificial Intelligence
3.Template matching

Enhancement Analysis Classification Recognition

Segmentation

Grade
Invasiveness
Normal
Malignant
63
IMAGE CLASSIFICATION/RECOGNITION
(APPLICATIONS)

Source: Matthew S. Brown, Computer-aided Lung Nodule Detection in CT: Results of Large-Scale Observer
Test1, Academic Radiology, Volume 12, Issue 6, June 2005, Pages 681-686, ISSN 1076-6332. 64
 LAST BUT NOT LEAST…
IMPROPER USE OF
IMAGE PROCESSING

ORIGINAL IMAGE

PROPER USE OF
IMAGE PROCESSING

65 65
 FURTHER READING
1. Gonzalez RC, Woods RE, “Digital image processing”, Prentice Hall; 2002
2. Pratt K. William, Digital Image Processing, John Wiley & Sons, 2001
3. Martinez W.L., Martinez A.R., Computational Statistics Handbook with MATLAB®,
Chapman & Hall/CRC, 2002
4. Young I. et al, Fundamentals of Image Processing, link
5. Housfield, Computed medical imaging, link
6. Cormack, Early two dimensional reconstruction and recent topics stemming from it, link
7. Mather S., The Principles of Diagnostic Imaging, link

66