Automated Segregation of Lymphoid and Myeloid Blasts in Acute Leukemia Cases Using a Deep Convolutional Neural Network
K. Anilkumar, V. J. Manoj, T. M. Sagi · Apple Academic Press eBooks · 2024
An abnormal rise in the count of immature white blood cells in the blood and bone marrow along with reduction in the count of normal blood cells may be an indication of leukemia. Classification of acute leukemia into lymphoblastic or myeloblastic is carried out by pathologists or hema-topathologists based on the morphological differences of the blast cells and with the support of cytochemical stains. Image-processing methods using machine learning can be used to classify leukemia by analyzing and processing images of bone marrow or blood smears. Such methods are simple, fast, and easy and can be used for assisting pathologists for a speedy classification of leukemia into various types and subtypes. This study performs automated classification of acute leukemia by segregating single blast cells into lymphoblasts and myeloblasts using ResNet50 , a pretrained 162 residual deep convolutional neural network , without the involvement of any image segregation and hand-crafted feature extraction. The proposed work successfully classified acute leukemia cells into lymphoblasts and myeloblasts with a classification accuracy of 94.9% using the pretrained network ResNet50. This chapter established that preprocessing techniques such as filtering, image enrichment, image segregation, and feature extraction are not necessary for the classification of leukemia smear images by deep learning methods. Classification of acute leukemia alone is considered in the study and classification of other types of leukemia can be done in future works.