LEUKEMIA CLASSIFICATION USING AN OPTIMIZED HYBRID MODEL: CNN, RESNET, DNN, AND XGBOOST IN PERSPECTIVE

Leukemia is a fatal cancer affecting individuals of all ages, primarily involving abnormal increases in immature white blood cells that impair bone marrow and blood function. Accurate and timely diagnosis is essential but current manual microscopic analysis is slow, error-prone, and challenged by the similarity between healthy and leukemic cells. This study proposes a novel Federated Learning (FL)-based Deep Learning (DL) framework for leukemia detection, ensuring privacy-preserving training across distributed clients. Input images are first denoised using an Adaptive MedianFilter (AMF), followed by Scribble2label-based cell segmentation and image augmentation.Features are then extracted and classified using a Dense Convolutional Network (DenseNet), optimized through a proposed Fractional Mayfly Optimization (FMO) method, which integrates the Mayfly Algorithm and Fractional Calculus.Local model updates are aggregated at the server via RV coefficient-weighted averaging.The