Conference PaperPDF Available

Detection of Blood Count Related Diseases from Blood Cell Images

Authors:
  • Director, IIIT Kottayam, Kerala, India Institute of National Importance

Abstract

Medical image analysis plays a major role in providing quality health care [3J. Better imaging techniques have enabled effective diagnosis of various diseases using autothresholding and watershed algorithm. Automation of the diagnostic processes is essential because the manual methods require considerable amount of time [4J, effort and care, besides being prone to errors. It also facilitates test data collection for the people in remote areas, where expert physicians are not widely available. Aim of medical image analysis is to develop systems that are capable of processing medical images required for diagnosis. The primary purpose of medical image analysis is to extract relevant information from images in order to facilitate unambiguous detection of abnormalities. Such systems can also be used for visualization. Computer based descriptions are often more consistent than those derived by human observers. The descriptions can include shape, color, pattern , texture, and other image features. The increasing availability of computing power and appropriate modeling techniques have enabled rapid development of Dr. R. V. Dharaskar Prof, Head Dept. of CSE GHRCE, Nagpur Mr. Jai M. PatH III SEM M.Tech (C.S.E.) RCERT. Chandrapur medical systems for quantitati e image anal~"'S that support disease detection , and medical education. A wide range of task in me analysis includes image enhanceIl:l!fB segmentation, noise removal, pattern detectiCi depending upon the requirement. For example MRI image segmentation helps in diagnosis. bI sometimes the low contrast MRI images nea enhancement before being used for diagnosit while the ultrasound fetal images would need • be analyzed for textures in order to detenn:a lung maturity. Patterns in bone images 211I recognized for forensic applications as well JI for the determination of age, study of age relall: bone developments and diseases. Angiogram, the X-ray image of netwC'l blood vessels needs to be processed :11 suppressing the shadows created by bon€s. i order to enable correct diagnosis. The images i3 skin moles need processing to eXInl descriptions that art aid the diagnosis ,I] melanoma, a skin cancer. Similarly segmentana of sputum images helps in the diagnosis of lUll cancer. Range of diseases caused by disordeTI i blood can be diagnosed with the help of bloo smears. Here, we are working with color inlag.c of blood smears to detect and classify ",-hil blood cells.
Detection of Blood Count Related Diseases from Blood Cell Images
Mr. Lekhraj Vilhekar
III SEM M. Tech
(C.S.E.) GHRCE,
Nagpur
lekh. vilhekar@ gmail.com
Dr. V.M. Thakare
Prof
&
Head, PGDCS,
Amravati University,
Amravati
Medical image analysis plays a major role in
providing quality health care [3J. Better imaging
techniques have enabled effective diagnosis of
various diseases using autothresholding and
watershed algorithm. Automation of the diagnostic
processes is essential because the manual methods
require considerable amount of time [4J,effort and
care, besides being prone to errors. It also
facilitates test data collection for the people in
remote areas, where expert physicians are not
widelyavailable.
Keywords: Authothresholding, Watershed
Algorithm.
Aim of medical image analysis is to
develop systems that are capable of processing
medical images required for diagnosis. The
primary purpose of medical image analysis is to
extract relevant information from images in
order to facilitate unambiguous detection of
abnormalities. Such systems can also be used for
visualization. Computer based descriptions are
often more consistent than those derived by
human observers. The descriptions can include
shape, color, pattern , texture, and other image
features. The increasing availability of
computing power and appropriate modeling
techniques have enabled rapid development of
Dr. R. V. Dharaskar
Prof, Head Dept. of CSE
GHRCE, Nagpur
Mr.
J
ai M. PatH
III SEM M.Tech
(C.S.E.) RCERT.
Chandrapur
medical systems for quantitati e image anal~"'Si
that support disease detection ,
and medical education.
A wide range of task in me
analysis includes image enhanceIl:l!fBl
segmentation, noise removal, pattern detectiCi
depending upon the requirement. For example
MRI image segmentation helps in diagnosis.
bI
sometimes the low contrast MRI images
nea
enhancement before being used for diagnosit
while the ultrasound fetal images would need •
be analyzed for textures in order to detenn:a
lung maturity. Patterns in bone images
211I
recognized for forensic applications as well
JI
for the determination of age, study of age relall:l
bone developments and diseases.
Angiogram, the X-ray image of netwC'lil
blood vessels needs to be processed
:11
suppressing the shadows created by bon€s.
i
order to enable correct diagnosis. The images
i3
skin moles need processing to eXInll
descriptions that art aid the diagnosis
,I]
melanoma, a skin cancer. Similarly segmentana
of sputum images helps in the diagnosis of
lUll
cancer. Range of diseases caused by disordeTI
i
blood can be diagnosed with the help of bloo
smears. Here, we are working with color inlag.c
of blood smears to detect and classify
",-hil
blood cells.
1.1
Image
Acquisition
This is the process of capturing images
xblood smears. In this case, color images are
.:apmred using a digital camera mounted on a
Dk."TOscope.
U
Image
Segmentation
This step estimates the shape of WBC and
hi
efficiency of the subsequent stages depends on
16Ii::s...
In this stage, WBC's are extracted from the
-.:kground and other constituents of blood such as
IBC's,
plasma, platelets and cell fragments. Further
ii:Rinction between nucleus and cytoplasm of each
.::dl
is accomplished.
Features extracted from segmented cells
Be
generally shape based, color based and
1I::I.tUre
based. Shape based features include
o.:centricity of the nucleus, eccentricity of the
~plasm, area ratio between the cytoplasm and
:be
nucleus, number of nucleus lobes etc. Color
!Bsed features used are average red, blue, green
axnponent for nucleus and cytoplasm. While
~Us~ textures based features are energy,
CBIlropy,~elatiOn etc. It has been observed
dJOiIl
shape b~sed features are most important
z-e, )
Based on the features extracted, a set of
i::amre
vectors, representing all 5 types of
1rrBC's is created. This set is used to train the
dassification model. After training, an unseen
sample of WBC can be classified as one of the 5
types. For classification purpose widely used
classifiers are neural networks, support vector
machines and Bayesian classifier.
Most techniques proposed so far are
sensitive to the right selection of some
parameters such as image acquisition conditions,
threshold selection, and initial contour. Also
some of the techniques assume circular shape for
white blood cells, which is not true in most
cases. A robust technique for segmentation is
proposed that performs further segmentation
between cytoplasm and red blood cells. There is
also a two stage segmentation scheme that
enables us to distinguish the cytoplasm and
nucleus of WBC from the input image of a blood
smear and requires no manual interaction for
parameter tuning.
Segmentation is achieved using
Autothrosholding and watershed algorithm.
Each sub-image is separately processed
for this stage. It has been observed that given the
gray scale image of any cell, dark regions
correspond to the nucleus, bright regions
correspond to background and intermediate
regions correspond to cytoplasm and RBC' s. So
thresholding the image into three classes,
separate cell structures from one another, except
cytoplasm and RBC's. We proposed automatic
threshold selection proposed by Otsu. In this
method, optimal thresholds Tl and 1'2 are
selected by maximizing interclass variance
between dark, gray and bright regions.
The watershed algorithm is a fundamental image
segmentation tool in mathematical morphology.
The watershed transform is based on an analogy
with topographic reliefs. An image can be
thought of as a three dimensional relief with the
grayscale value at each point corresponding to
height. Imagine that the relief has a point. Once
the relief has become completely covered by
water, we endup with a structure with several
barriers or dams on it. These dams represent the
watershed lines and serve to separate the
"catchment basins "of the relief. One of the main
advantages of the watershed transform as a
segmentation tool is that the segment boundaries
it produces are closed.
We apply watershed algorithm on the
binary image of the cytoplasm. As a result of
declustering we obtain a cluster which belongs
to the cytoplasm. The advantage of using the
watershed algorithm is that the contour
information is not lost. In some cases, due to
over segmentation, cytoplasm is divided into
several clusters, so in order to get binary mask
for cytoplasm we need to merge these clusters.
Proposed system will enhance the
working of medical system just by
diagnosing blood diseases with full
automation. Using various algorithms, it is
required to first detect edges in the image
and segment the image to analyze .•
various cells to detect the disease.
Image
Enhance
ment
Edge
Detection
Image
Segmenta
tion
Counting
of cells
Proposed system requires
identification froin the blood cell
So, image consists of various
fj
[14]
like shape, color, pattern, and .-__
etc. It is required to first find the ' •••••
contents of the image, if the linage
:S ••
available with desired resolution .•• "
image needs to be enhanced... .~
enhancement, the various edges:r
111I:
cell will be detected to differential~
::i::nmm
the background. The various alg;~11
like Watershed [15] and autotbres.b=L:lI:~
,,,; ::5
used for the segmentation of the
JIDIBI8I!!I=C
The various cells can be counted
iIDml
each of the segments and after total
.i'_'@'lIg
the result is analyzed [17].
_ :ills
efficiency with the existing
" ilhllls
to give better results than the
~: In medical image analysis , until
d,
work has been done over
C dation of various blood cells
, C ,
like white blood cells [7], red
--.. cells, peripheral cells, Para
, Z·ljIric
cells, parasite cells [8] with
MZ -ithm
[9], autothresholding [9], and
•• algorithm [10] and morphological
• "'lIs[11]. Various neural network
·""es
have been used for the detection
~ aI!es of cells with different algorithms
P3l-
.-'i.
I
L
~~.wadSWOrth.orgichemheme
e1microscope/
1.
ez, Digital Image Processing,
Addison-Wesley, 1993.
3.
K.S. Kim, P.K. Kim, J.J.Song, and
Y.C. Park,( 2002), "Analyzing blood
cell image to distinguish its
abnormalities", in Proc., ACM
international conference on
Multimedia.
4. BAa, H.F. Den, H.H. Gelsema,
Smeulders, "Automated while blood
cell classification revisited ". Medical
Informatics 12, 1(1997),23-31.
5. Libo Zeng, Hong Zeng, Ningning
guo, " Precise segmentation of white
blood cells by using multispectral
imaging analysis techniques "., at first
international conference on Intelligent
networks and intelligent systems, 491-
494.
6. Fabio Scotti, "Robust Segmentation
and Measurements Techniques of
White cells in blood microscope
images." 24-27 April 2006 at
IMTC2006 Instrumentation and
Measurement Technology
Conference, Sorrento, Italy, 43-48.
7. Q. Liao and Y. Deng, " An accurate
segmentation method for white blood
cell images", in IEEE International
symposium on Biomedical Imaging.,
2002.
8. Dumpster, Di Ruberto,
"Morphological Processing of
Malarial Slide Images", Mat lab DSP
Conference 1999, Nov, 16, Espoo,
Finland .
9. P. Bamford and B. Lovell, "Method
for accurate unsupervised cell nucleus
segmentation", in Proc., of the
Engineering in Medicine and Biology
society conference, 2001.
10. Neelam Sinha, A.G. Ramkrishnan,
"Blood cell segmentation using EM
algorithm" .
11. Andrew Dempster, Shahid khan, Bill
Jara, "Segmentation of blood images
using morphological oprators"., 397-
400.
12. BAMFORD, LOVELL, "Method for
accurate unsupervised cell nucleus
segmentation". In EMBS Conference
(2001), pp.2704-2708
13. Ruihu Wang, Brendan McCane, "Red
blood cell classification through
Depth Map and surface feature ", at
International symposium on computer
science and computer technology,
2008,339-341.
14.Harms, H., Gunzer, D., Aus H.M.,
"Combined local color and texture
analysis of stained cells ", Computer
Vision, Graphics and image
processing, 33, 364-376, 1986.
15. S.Beucher, "The Watershed
Transformation Applied to Image
Segmentation".
16. S.S. S. Poon, R.K. Ward, and B.
.Palcic, "Automated image detection
and segmentation in blood smears",
Cytometry, vol. 13,pp.766-774, 2002.
17.D.S. Serbouti et aI., "Image
segmentation and classification
methods to detect leukeniia's", in
Proc. International conference of
IEEE engineering in Medicine and
Biology society, 1991.
18. Nipon Theera-Dmpon and Paul D.
Gader, " System Level training of
neural networks for counting white
blood cells.", IEEE transactions on
systems, MAN, and cybernetics.-
PART C : Applications and reviews,
VpI., 32 No.1, February 2002, 48-53.
19.Andrew Dempster, Izzet Kale,
"Malaria Parasite Detection in
Peripheral Blood Images".
20. Marcin Iwanowski and Pierre Soille ,
" Morphological refinement of an
image segmentation".
21. Blood,
https://courses.stu.qmul.ac.uk/smd/kb/
microanatomy/blood/.
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
This paper investigates the possibility of computerised diagnosis of malaria and describes a method to detect malaria parasites (Plasmodium spp) in im- ages acquired from Giemsa-stained peripheral blood samples using conven- tional light microscopes. Prior to processing, the images are transformed to match a reference image colour characteristics. The parasite detector utilises a Bayesian pixel classifier to mark stained pixels. The class conditional prob- ability density functions of the stained and the non-stained classes are esti- mated using the non-parametric histogram method. The stained pixels are further processed to extract features (histogram, Hu moments, relative shape measurements, colour auto-correlogram) for a parasite/non-parasite classi- fier. A distance weighted K-nearest neighbour classifier is trained with the extracted features and a detailed performance comparison is presented. Our method achieves 74% sensitivity, 98% specificity, 88% positive prediction, and 95% negative prediction values for the parasite detection.
Conference Paper
Full-text available
This paper describes a method to improve a given segmentation result in order to produce a new, refined and more accurate segmented image. The method consists of three phases: shrinking of the input partitions, filtering of the input imagery leading to a mask image, and expansion of the shrunk partitions within the filtered image. The concept is illustrated for the enhancement of a land cover data set using multispectral satellite imagery.
Article
Full-text available
Neural networks (NNs) that are trained to perform classification may not perform as well when used as a module in a larger system. We introduce a novel, system-level method for training NNs with application to counting white blood cells. The idea is to phrase the objective function in terms of total count error rather than the traditional class-coding approach because the goal of this particular recognition system is to accurately count white blood cells of each class, not to classify them. An objective function that represents the sum of the squared counting errors (SSCE) is defined. A batch-mode training scheme based on back-propagation and gradient descent is derived. Sigma and crisp counts are used to evaluate the counting performance. The testing results show that the network trained to minimize SSCE performs better in counting than a classification network with the same structure even though both are trained a comparable number of iterations. This result is consistent with the principle of least commitment of D. Marr (1982)
Article
A simple technique which automatically detects and then segments nucleated cells in Wright's giemsa-stained blood smears is presented. Our method differs from others in (1) the simplicity of our algorithms; (2) inclusion of touching (as well as nontouching) cells; and (3) use of these algorithms to segment as well as to detect nucleated cells employing conventionally prepared smears. Our method involves: (1) acquisition of spectral images; (2) preprocessing the acquired images; (3) detection of single and touching cells in the scene; (4) segmentation of the cells into nuclear and cytoplasmic regions; and (5) postprocessing of the segmented regions. The first two steps of this algorithm are employed to obtain high-quality images, to remove random noise, and to correct aberration and shading effects. Spectral information of the image is used in step 3 to segment the nucleated cells from the rest of the scene. Using the initial cell masks, nucleated cells which are just touching are detected and separated. Simple features are then extracted and conditions applied such that single nucleated cells are finally selected. In step 4, the intensity variations of the cells are then used to segment the nucleus from the cytoplasm. The success rate in segmenting the nucleated cells is between 81 and 93%. The major errors in segmentation of the nucleus and the cytoplasm in the recognized nucleated cells are 3.5% and 2.2%, respectively. © 1992 Wiley-Liss, Inc.
Combined color and texture techniques provide new, useful features for blood cell analysis. After scene segmentation, texture line following programs define the borders of textons, the smallest texture grains. Size, color, and shape features are then calculated for each texton. The final feature data set includes: (1) The nucleus and cytoplasm features: size and shape, measurements of the texture lines from the green TV image, average CIE-PAL color coordinates and a color code. (2) The new texton features: total number of texton pixels with a specific color code, mean texton radius, size for each color code, and texton shape factors.