International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT) - 2016

Detection of Blood Cancer in Microscopic Images of

Human Blood Samples: A Review
M.Saritha Prakash.B.B
Department of Electronics & Communication Engineering Department of Electronics & Communication Engineering
Jain College of Engineering Jain College of Engineering
Belgaum , India Belgaum, India
sarithamiryala5@gmail.com

Sukesh.K Shrinivas. B
Department of Electronics & Communication Engineering Department of Electonics & Communication Engineering
Jain College of Engineering , Belgaum, India Jain College of Engineering, Belgaum ,India

Abstract— For the fast and cost effective production of leukemia caused due to lymphoid cells is called lymphocytic
patient diagnosis, various image processing techniques or or lymphoblastic leukemia and if it is found in the myeloid
software has been developed to get desired information from cells, it is called myelogenous or myeloid leukemia [3].
medical images. Acute Lymphoblastic Leukemia (ALL) is a type Leukemia is grouped in 2 ways: Acute or chronic which is
of leukemia which is more common in children. The term grouped on the basis of how fast the cells grow. Lymphoid or
‘Acute‘means that leukemia can progress quickly and if not
treated may lead to fatal death within few months. Due to its non
myeloid is another type of leukemia which is categorized
specific nature of the symptoms and signs of ALL leads wrong depending on the type of white blood cell that has turned into
diagnosis. Even hematologist finds it difficult to classify the leukemia[2],[3].[4]. In acute leukemia, the abnormal blood
leukemia cells, there manual classification of blood cells is not cells are usually immature blasts (young cells) that do not
only time consuming but also inaccurate. Therefore, early function properly. These cells grow quickly. Acute leukemia
identification of leukemia yields in providing the appropriate quickly gets worse unless it is treated immediately. In chronic
treatment to the patient. As a solution to this problem the system leukemia, young blood cells are present, but mature,
propose individuates in the blood image the leucocytes from the functional cells are also produced. In chronic leukemia, blasts
blood cells, and then it selects the lymphocyte cells. It evaluates grow slowly. It takes longer for the disease to get worse.
morphological index from those cells and finally it classifies the
presence of leukemia. In this paper a literature review is been
conducted on various techniques used for detecting cancer cells. The four major forms of leukemia are

1. Acute lymphoblastic leukemia (ALL)

Keywords— Blood Cancer, Image Segmentation, Acute 2. Acute myelogenous leukemia (AML)
Lymphoblastic Leukemia (ALL), Leukemia, Morphological 3. Chronic lymphocytic leukemia (CLL) and
Analysis. 4. Chronic myelogenous leukemia (CML)
I. INTRODUCTION (HEADING 1)
The infected cells can be observed in microscopic
Leukemia is a type of cancer which affects the white image, manually by a trained expert where the unique features
blood cells. Blood consists of plasma, and three different types are visually observed and the classification of type of cancer is
of cells (White Blood Cells, Red Blood Cells and Platelets) done. The variety of features and often unclear images results
and each of these performs particular task. White blood cells in missing of data which can be vital indicator to differentiate
help the body to fight against infection and diseases. Red the type of cancer, therefore identification task becomes
blood cells carry oxygen from the lungs to the body's tissues difficult. [2], [3], [4]. In the following section work done by
and vice versa. Platelets help to form blood clots and control different researchers has been presented.
bleeding[1]. A body of the person who is suffering from
leukemia produces too many blood cells of particular type
than another which results in abnormalities in the cells. These II. RELATED WORK
abnormal cells look different than normal blood cells and do
not function properly (usually white blood cells)[3]. They also Various image processing techniques has been
interfere with other blood cells, usually red blood cells and developed by researchers to detect the blood cancer in
platelets. biomedical images of human blood samples. Some of them are
Two types of abnormal white blood cells can turn [5], uses thresholding technique in determining the ratio of
into leukemia: lymphoid cells and myeloid cells. When blood cells for cancer cells detection. In this paper image

978-1-4673-9939-5/16/$31.00 ©2016 IEEE

596
processing techniques has been used to count the number of The output of adaptive threshold algorithm were given to
blood cells in the biomedical image. With this counted value KFCM clustering and the fuzzy image with fuzzy
of blood cell, the ratio of blood cell for leukemia is calculated. boundaries are obtained. Adaptive and KFCM both can be
The original image which is converted to grayscale image for together used for the medical images and low intensity images
which a threshold value of intensity is set inorder to and resultant image has better accuracy and quality. The
differentiate WBC toRBC (thresholding converts a gray scale results of adaptive thresholding are better than global
image to binary image). If results are not satisfied the process thresholding.
is repeated by setting a new threshold value. The results
acquired using thresholding technique shows that the ratio of [13], in this paper image segmentation is done using wavelet
RBC and WBC for normal image to the abnormal image has transform and its results are compared with the conventional
different range of ratio. For normal images the ratio is 0 to 0.1 segmentation technique and the results obtained using wavelet
whereas for abnormal images its ratio range is 0.2 to 2.5 for transforms are better than conventional methods.
ALL and 0 to 14 for AML. The disadvantage of this technique
is setting of proper threshold value would be difficult and time [14], has segmented the image using K Means clustering and
consuming. neural networks. The results are analyzed by comparing
various error metrics and time complexity of the both the
[6], uses two methods of image segmentation i.e. thresholding algorithms. Image segmentation done using feed forward
and watershed to detect the cancer infected cells. The neural networks has less noise and a very slow convergence
segmentation accuracy using thresholding and watershed rate. In this research K-means clustering and artificial neural
techniques are 81.24% and 85.27% respectively. From the networks (ANN) are studied to obtain a set of algorithms
results watershed segmentation of biomedical image has better which can be combined in order to achieve a better
accuracy and quality. performance in image segmentation. A comparison has been
done to find out the best algorithm for image segmentation.
[7], thispaper proposes the use of morphological analysis of
leukemic infected image thereafter segmentation has been [15], has propose k-means and improved watershed
done which provides 2 enhanced images of cytoplasm and segmentation algorithm for medical image segmentation. The
nuclei regions. Here bimodal thresholding has been used conventional watershed algorithm has an advantage of
,where two threshold values of intensities was given to get the complete division of image but the disadvantage is over-
images of cytoplasm and nuclei and boundary of infected cell segmentation and sensitivity. The improved watershed
is traced and unwanted objects in the surrounding is segmentation algorithm makes use of an automated
filtered.From the results 98.33% of overall sufficient thresholding. This segmentationreduces theover-segmentation
segmentation ratio of was obtained. and also false edges. In this paper proposed method produces
segmentation maps which have 92% fewer partitions than the
[9],has proposed segmentation framework that consists of an segmentation maps produced by the conventional watershed
integration of several digital image processing algorithm.
algorithmswhich is called Zack Algorithm. A sample of
twenty microscopic blood images were tested and obtained III. RESEARCH METHODLOGY
92% accuracy for nucleus segmentation and 78% for
cytoplasm segmentation with this results it is possible to From the literature review the following are steps to
extract the nucleus and cytoplasm region in a WBC image be followed for the automation of medical image recognition,
sample. shown in the figure 3.1

[10], for enhancement and segmentation of image, automatic

Otsu’s threshold has been proposed in this paper.kNN
classifier has been utilized to classify blast cells from normal
lymphocyte cells. The system can be applied to 108 images
which are available in public image dataset for the study of
leukemia. This method has given 93% accuracy. The k-
Nearest- Neighbours (kNN) is method of classifying non-
parametric variables. This method is simple but very effective
which can also be used todifferentiate blast cells from normal
white blood cells.

[11], used adaptive thresholding technique using kernel fuzzy

clustering method (KFCM). Instead of having single
Figure 3.1 Sequence of steps for automatic blood sample
threshold value, a threshold value which changes dynamically
image recognition
with the image needs is used called adaptive thresholding.

597
Research methodology follows: extraction is essential for the classification of desired objects.
[23]. Feature Extraction is an essential important pace in the
1. Microscopic Image Acquisition construction of any pattern classification which intends to
extract the relevant information that differentiates each [8]
Microscopic image of blood cells has to be obtained
from nearby hospital with sufficient magnification. • Geometrical Features consists of geometrical parameters like
area, radius, perimeter, symmetry, boarder, concavity,
2. Image Enhancement ( Preprocessing) compactness, solidity, eccentricity, elongation, form factor.

The medical image obtained may contain noise due to • Texture Features consists of homogeneity, energy,
excessive staining. The image has to be denoised using any of correlation, entropy contrast, angular second momentum.
the image denoising technique which improves the quality of
the image but the image may appear as slightly blurred. The • Color Features consists of the RGB color spaces will be
background of the blood sample image has to be excluded transformed into HSV color spaces. Their mean color values
because the Region of Interest is white blood cells. This will be obtained.
preprocessing of image, like noise reduction and enhancement
will make the edges of the white blood cells sharper in the • Statistical Features consists of the mean value, variance,
image [22]. skew ness, kurtosis of the histograms of the image matrix and
the gradient matrix for RGB or HSV color space.
3. Image Segmentation
Based on [19], ALL is small, blast cells are uniform,
The aim of the image segmentation is partition of an cytoplasm is scanty, round and usually contains single nucleoli
image into multiple segments. In this particular case the image inside nucleus. While in AML, the blasts are larger and
segmentation partitions the white blood cell (WBC) from irregular form and usually multiple nucleoli with the presence
RBCs and plasma in the blood sample image. Image of Auer rode. [24] said that, the WBC appears rather darker
segmentation technique is a very hot area of research in the than the background while red blood cell (RBC) appears in an
field of Image Processing. A lot of research work as been done intermediate intensity level. [25] indicates that white cells are
on image segmentation and there are different segmentation the darker elements in images with RBC appear to be pale.
algorithms but no single algorithm will be suitable for Platelets are much smaller than white and red cells.
different types of images. Therefore as a result an algorithm
designed for one particular image cannot be applied for other 5. Image Classification
type of image [22]. Therefore developing a single
segmentation technique for different images still remains a Depending on the features extracted from the above step
changing job. An automatic image Segmentation technique the classifier classifies a cell as normal cell or cancer affected
which gives the accurate results and which consumes less time cell i.e. blast cell. This is done by comparing some of the
especially for medical images can help in saving life by early features like geometric, statistical, texture and size ratio from
detection of disease and early treatment. regions obtained in the segmentation process with standard
feature. Then the results are analyzed to identify the types and
Work done in medical image segmentation by various subtypes of acute leukemia[25][24].
researchers and various techniques has been discussed in Before you begin to format your paper, first write and save
chapter II. Abnormal immature white blood cells are called the content as a separate text file. Keep your text and graphic
blasts. The Proposed method should be able to differentiate files separate until after the text has been formatted and styled.
the blast cell from the normal cells. For determining whether a Do not use hard tabs, and limit use of hard returns to only one
cell is a blast or no, lymphocytes and myelocytes of blood return at the end of a paragraph. Do not add any kind of
only has to be considered and others like neutrophils, pagination anywhere in the paper. Do not number text heads-
basophils and eosinophils can be barred from the image. the template will do that for you.
When the blast cells are identified it can proceeded for the
next step and sub images which contains nucleus only has to IV. DISCUSSION
be focused which reduces the errors because there can be Some of the issues which have to be resolved about blood
similar color scales in WBCs with other blood particles [23]. cells are blood sample itself. Due to environmental pressures
cells get deformed to arbitrary shapes, [16] Because of
4. Image Feature Extraction diseases overlapped cells may get joined. This problem of
overlapping can be overcome by watershed method.[6],[17].
Feature extraction is the technique which extracts the Another issue regarding blood cells is collection of data
desires features from the pre-processed images which contains because the collecting more samples is a not an easy task.
of different abnormalities. The different features of the image Filtering and elimination of unwanted objects from the cell
can be size, shape, composition, location etc. Feature region. All the above mentioned issues have to be considered

598
while designing a system. Therefore an automated image Image Processing” in Shulin Li (ed.), Biological Procedures Online,
Volume 11, Number 1 to the author(s) 2009
analysis system has to be developed to measure morphologic
parameters and evaluation of cervical cells.
[10] Ms. Minal D. Joshi, Prof. Atul H. Karode, Prof. S.R.Suralkar, “White
Blood Cells Segmentation andClassification to Detect Acute Leukemia”
in International Journal of Emerging Trends & Technology in Computer
Science (IJETTCS), Volume 2, Issue 3, May – June 2013
V. CONCLUSION
[11] Tara.Saikumar, B.K.Anoop, P.S.Murthy in “Robust Adaptive Threshold
This research contains detection of blood cancer cells Algorithm based on Kernel Fuzzy Clustering on Image segmentation” in
Natarajan Meghanathan, et al. (Eds): ITCS, SIP, JSE-2012, CS & IT 04,
and classification of the types of leukemia from microscopic pp. 99–103, 2012. © CS & IT-CSCP 2012
image samples using image processing. The proposed method [12] Michael Barnathan,“Mammographic Segmentation Using
extracts the features in microscopic images by examining WaveCluster”, open access algorithms ISSN 1999-4893, Algorithms
changes on various parameters like texture, geometry, colors 2012.
and statistical analysis input. The system should have high [13] Xu Gongwen, Zhang Zhijun, Yuan Weihua, XuLi Na, “On Medical
reliability, accuracy and efficacy, less processing time, smaller Image Segmentation Based on Wavelet Transform” in IEEE , ISBN
978-1-4799-4262-6, June 2014
error, less cost and must be robust. Early identification of
leukemia yields in providing the appropriate treatment to the
[14] Chinki Chandhok “A Novel Approach to Image Segmentation using
patient. Artificial Neural Networks and K-Means Clustering” in Chinki
Chandhok / International Journal of Engineering Research and
Applications (IJERA) ISSN: 2248-9622 ,Vol. 2, Issue 3, May-Jun 2012,
References pp. 274-279

[15] Ng, HP ong,S.H; Fooong, K.W.C in “Medical Image Segmentation

[1] C.R., Valencio, M.N., Tronco, A.C.B.,Domingos, C.R.B., “Knowledge Using K-Means Clustering and Improved Watershed Algorithm” IEEE ,
ExtractionUsing Visualization of Hemoglobin Parameters to Identify ISBN 1-4244-0069-4,2006
Thalassemia”,Proceedings of the 17th IEE Symposium on Computer
Based Medical Systems, 2004, pp.1-6.
[16] N., Ritter, J., Cooper, “Segmentation and Border Identification of Cells
in Images of Peripheral Blood Smear Slides”, 30th Australasian
[2] R., Adollah, M.Y., Mashor, N.F.M, Nasir, H., Rosline, H., Mahsin, H., Computer Science Conference, Conference in Research and Practice in
Adilah, “Blood Cell Image Segmentation: A Review”, Biomed2008, Information Technology, Vol. 62, 2007, pp. 161-169
Proceedings 21, 2008, pp. 141-144.

[17] I., Cseke, “A Fast Segmentation Scheme for white Blood Cell Images”,
[3] N., Ritter, J., Cooper, “Segmentation and Border Identification of Cells IEEE, 1992, pp. 530-533.
in Images of Peripheral Blood Smear Slides”, 30thAustralasian Computer
Science Conference, Conference in Research and Practice in Information
Technology, Vol. 62, 2007, pp. 161-169. [18] A., Khashman, E., Esam Al-Zgoul, “Image Segmentation of Blood Cells
in Leukemia Patients”, Recent Advances in Computer Engineering and
Applications, ISBN: 978- 60-474-151-9, 2009, pp. 104-109.
[4] D.M.U., Sabino, L.D.F., Costa, L.D.F., E.G., Rizzatti, M.A., Zago, “A
Texture Approach to Leukocyte Recognition”, Real Time Imaging, Vol.
10, 2004, pp. 205-206. [19] N.I., Birndorf, J.O., Pentecost, J.R., Coakley, K.A., Spackman, “An
Expert System to Diagnose Anemia and Report Results Directly on
[5] Abdul Nasir, Mustafa N, Mohd Nasir, “ Application of Thresholding in
Hematology Forms”, Computers and Biomedical Research, Vol. 29,
Determing Ratio of Blood Cells for Leukemia Detection” in the
1996, pp. 16-26.
Proceedings of International Conference on Man-Machine system
(ICoMMS) oct 2009 ,Batu Ferringhi, Penang, Malaysia.
[20] S., Mohapatra, D., Patra, S., Satpathi, “Image Analysis of Blood
Microscopic Images for Leukemia Detection”, International Conference
[6] Bhagyashri G Patil, Prof. Sanjeev N.Jain , “Cancer Cells Detection
on Industrial Electronics, Control and Robotics, IEEE, 2010, pp. 215-
Using Digital Image Processing Methods” in International Journal of
219.
Latest Trends in Engineering and Technology”.

[21] T. Markiewicz, S. Osowski, B., Marianska, L., Moszczynski,

[7] ADNAN KHASHMAN, ESAM AL-ZGOUL, “Image Segmentation of
Blood Cells in Leukemia Patients” in RECENT ADVANCES in
“Automatic Recognition of the Blood Cells of Myelogenous Leukemia
Using SVM”, Proceedings of International Joint Conference on Neural
COMPUTER ENGINEERING and APPLICATIONSISBN: 978-960-
Networks, Montreal , Canada, July 31 – August 4, 2005, pp. 2496-2501
474-151-9

[22] N., H., A., Halim, M., Y., Mashor, R., Hassan, “Automatic Blasts
[8] FAUZIAH KASMIN, ANTON SATRIA PRABUWONO, AZIZI
Counting for Acute Leukemia Based on Blood Samples”, International
ABDULLAH “DETECTION OF LEUKEMIA IN HUMAN BLOOD
SAMPLE BASED ON MICROSCOPIC IMAGES: A STUDY” , Journal of Research and Reviews in Computer Science, Vol. 2, No. 4,
Journal of Theoretical and Applied Information Technology 31st August 2011, pp. 971-976.
December 2012. Vol. 46 No.2.
[23] Athira Krishnan, Sreekumar K, “ A Survey on Image Segmentation and
Feature Extraction Methods for Acute Myelogenous Leukemia Detection
[9] Farnoosh Sadeghian, Zainina Seman, Abdul Rahman Ramli, Badrul in Blood Microscopic Images” , (IJCSIT) International Journal of
Hisham Abdul Kahar, and M-Iqbal Saripan, “A Framework for White Computer Science and Information Technologies, Vol. 5 (6) , 2014, 7877-
Blood Cell Segmentationin Microscopic Blood Images Using Digital 7879

599
[24] A.Sindhu1, S.Meera2 “ A Survey on Detecting Brain Tumorinmri [25] Niranjan Chatap, Sini Shibu “Analysis of blood samples for counting
Images Using Image Processing Techniques” , International Journal of leukemia cells using Support vector machine and nearest neighbor”, e-
Innovative Research in Computer and Communication Engineering. Vol. ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 16, Issue 5, Ver. III (Sep –
3, Issue 1, January 2015. Oct. 2014), PP 79-87 www.iosrjournals.org..

600