Detection and Classification of Acute Lymphoblastic Leukemia from Blood Smear Images Using Supervised Pattern Classifier

Abstract

Leukemia is a blood cancer that causes abnormalities in the lymphatic system of our body. It
is considered as one of the fatal diseases in our country. Among all the types of leukemia,
Acute Lymphoblastic Leukemia (ALL) is the fast growing disease. Hence, early detection of
this disease is needed in order to provide the required treatment to the victims. In this paper,
the detection and classification of ALL and its subtypes (L1, L2 and L3) is done by multiclass
SVM classifier based on the features extracted from nucleus and cytoplasm of the blast cell
(where abnormal leucocytes are known as ‘blast cells’). We tested our model on 100 images,
which contains 35 healthy cells, 37 L1 type cells, 9 L2 type cells and 19 L3 type cells. The
results acquired shows that SVM classifier obtained better results for two cells classification
since obtained accuracy of 2 class SVM classifier (Healthy, Malignant) is higher than the
accuracy achieved for 4 class SVM classifier with ECOC technique (Healthy, three ALL
subtypes).