Efficient Segmentation of Lymphoblast in Acute Lymphocytic Leukemia

Abstract

Microscopic examination of peripheral blood smears and bone marrow is the preliminary step for the diagnosis of several life-threatening diseases. Acute lymphocytic leukemia (ALL) is the most common disease in children that also needs an early diagnosis for on-time treatment as it spreads rapidly in the blood and forms immature lymphocytes. This might cause death in some weeks if left untreated. Manual methods in clinical laboratory being applied for the diagnosis of these diseases are inefficient and expensive, and the results are less accurate. A computer-aided system is the need of the day in which the most important step is segmenting the region of interest in blood or bone marrow for the detection and cure of the diseases which is the most challenging task. This study aims to propose a simple threshold-based segmentation technique by processing the S component of the HSV color space to segment the lymphoblasts in the bone marrow images of ALL patients. The technique was applied to 230 RGB bone marrow images having all the three types of ALL, i.e., L1, L2, and L3 resulted in the overall accuracy of 96.8%.