LEUKEMIA DETECTION AND CLASSIFICATION THROUGH DEEP LEARNING

Abstract

Every year, the medical field undergoes a modernization process and moves closer to increasingly automated technologies that support and enhance healthcare practices to be more accurate in their assessments and more productive in their treatments. A proper and timely diagnosis is necessary for the rehabilitation and treatment of leukemia, a type of cancer that can be lethal. Automated computer technologies for symptom analysis, diagnosis, and prediction have replaced traditional approaches. Modern techniques can assist patients in identifying life- threatening conditions like leukemia, a deadly illness and prevalent form of childhood cancer. In the proposed approach, leukemia subtypes are identified using Convolutional Neural Networks InceptionV3 using LIME (Local Interpretable Model-agnostic Explanations) Algorithm and Residual Convolutional Neural Network (ResNet-50). By incorporating the ResNet50 CNN architecture, the suggested model obtained the best accuracy of 98.84%.