Diagnosis of Leukemia Based on White Blood Cell Classification

Abstract

Blood diagnosis is based on a visual examination of blood smears, which is usually a time-consuming and error-prone process. In order to overcome this challenge, image processing techniques are required to assist the clinical decision-making process. Leukemia is a form of cancer, which is distinguished by the abnormal generation of immature white blood cells (WBC) known as tumors. Generally, leukemia will affect the white blood cells present in the bone marrow and/or blood. Hence, developing a prompt, safe, and accurate leukemia diagnosis is critical. White blood cells are often examined under a blood smear microscope for diagnosis purpose. On the other hand, numerous machine learning algorithms have been developed to diagnose different diseases, e.g., leukemia, and to provide a high number of misclassification error rates. The proposed research study has implemented a deep learning algorithm to classify the microscopic images for white blood count analysis. The WBC differential system is divided into two modules: detection and segmentation. The detection module initially evaluated the green bone smear pictures, detecting all WBCs in red blood cells, platelets, counts, and so on. The obtained cells are then fed into a separate module. The segmentation module was divided into two pieces. Numerous cells, including crushed cells and degraded new cells, are identified at the earliest stage of leukemia diagnosis. The WBCs may then be computed and distributed to multi-class differentiation stage by using a convolutional neural network [CNN] technique.