Distinguish normal white blood cells from leukemia cells by detection, classification, and counting blood cells using YOLOv5

Abstract

Leukemia is a group of blood cancers that usually begin in the bone marrow and result in high numbers of abnormal blood cells. Distinguishing normal white blood cells from leukemia cells plays a role in the diagnosis of blood diseases. Over the years, manual methods can detect, classify, and count blood cells that are labor-intensive slow, time-consuming, subject to error, and costly. An automatic system is applied in the processing detection through microscopic images that will be fast, cheap, and high accuracy without special needs of equipment for lab testing. In this paper, we propose a method of detection, classification, and counting to distinguish five types of normal white blood cells (WBCs) from abnormal white blood cells (leukemia) using the You Only Look Once version 5 (YOLOv5) to aid the diagnosis of blood diseases. After using the YOLOv5 algorithm, we manually labeled the blood cells, and the leukemia cells labeled are the largest with 619 cells, while the number of normal WBCs labeled is a basophil, 115 neutrophils, 23 eosinophils, 80 lymphocytes, and 73 monocytes, respectively. The tested result with the average accuracy of detection and classification of the blood cells is 93.0%.