Segmentation of White Blood Cells in Acute Myeloid Leukemia Microscopic Images: A Review

Abstract

Acute Myeloid Leukemia (AML) is a fast-growing leukemia caused by the rapid proliferation of immature myeloid cells. AML is a life-threatening disease if left untreated. Therefore, early detection of AML is crucial, maximizes the cure opportunities, and saves patients’ lives. Initial AML diagnosis is done by expert pathologists where blood smear images are utilized to detect abnormalities in WBCs. However, manual detection of AML is subjective and prone to errors. On the contrary, computer-aided diagnosis (CAD) systems can be an accurate diagnostic tool for AML and assist pathologists during the diagnosis process. Segmentation of White Blood Cells is the first step toward developing an accurate CAD system for AML. To date, WBC segmentation has several challenges due to several reasons such as different staining conditions, complex nature of microscopic blood images, and morphological diversity of WBCs. Current WBC segmentation techniques vary from conventional image processing methods to advanced machine learning and deep learning methods. This chapter discusses current segmentation methods as well as the potential solutions for improving automated WBC segmentation accuracy.