Automated Detection Leukemia Subtype Classification from Microscopic Images: A Convolutional Neural Network Approach

Abstract

Leukemia is a difficult type of blood cancer which can present in a combination of multiple types, that had individual cellular and genetic abnormalities. The prevalent forms of leukemia include Acute lymphocytic leukemia (ALL), Acute myeloid leukemia (AML), Chronic lymphocytic leukemia (CLL) and Chronic myeloid leukemia (CML). The evaluation of diseases, risk detection, and treatment planning all depend on the identification of specific genetic anomalies. This work describes a method for recovering Leukemia blood cancer cells applying blood microscopic images, identifying the subtypes of Leukemia. For detecting leukemia cancer using image processing techniques, that detect classification of leukemia blood cell subtypes. The proposed approach involves utilizing Convolutional Neural Network (CNN) to identify and categorize subtypes to leukemia blood cells based on microscopic images of human blood cells. Aimed to assess the expert Seven independent CNN models such as EfficientNetB7, ResNet-50, VGG19, ResNet101, DenseNet201, MobileNet and also built a CML-1(custom model) detect in the classification of leukemia. After exploit our proposed methodology, ResNet-50 model exhibited superior performance achieving 97.50% accuracy. The remaining models also displayed robust performance, with accuracy rates ranging from 95% to 97%. These finding imply that employing CNN methodologies for the automated identification of leukemia in microscopic blood images holds immense potential and presents substantial benefits in the realm of medical diagnostics.