Detection and Classification of Sub-Types of Acute Lymphocytic Leukemia Cells
Acute Lymphocytic Leukemia is a type of blood cancer which affects a group of white blood cells called
lymphocytes, causing the production of a vase amount of immature lymphocytes in the bone marrow. The microscope
examination for the morphological analysis of white blood cells is the initial screening step for the diagnosis of leukemia
and it is done by skillful pathologists. The problems in the diagnosis procedure arise due to the similarity of the malignant
cell structures between the other types of leukemia, causing the inconsistent prognosis result. Therefore, the accuracy of the
manual observation highly depends on the experience of the pathologists. In order to eliminate some drawbacks of manual
examination, a computer-assisted diagnosis system is proposed by performing the morphology analysis of the peripheral
blood smear microscope images. The methodology presented in this work consists of four main stages: (1) Pre-processing
(2) Segmentation (3) Feature Extraction (4) Classification. In pre-processing step, the input image is applied color contrast
adjustment, conversion to L*a*b color space and guided filtering for noise removal. Zack’s algorithm is used for segmenting
the leucocytes. Marker-controlled watershed is adopted for separation of touching cells. Segmentation of nucleus region is
achieved using K-Means clustering. Shape and textures features are extracted using Gray Level Co-occurrence
Matrix(GLCM) and histogram-based texture features methods. Finally, One-vs-all multi-Support Vector Machine is used to
classify the cells into L1,L2 and normal lymphocyte.
Index terms - Acute Lymphocytic Leukemia, Guided Filtering, GLCM, K-Means Clustering, Support Vector Machine.