Optimization of leukocytes using WOA-aided federated learning

Conventional practice in the medical industry dictates the use of manual methods for the diagnosis of immune system and bone marrow disorders. In order to diagnose these disorders quickly, the idea is to do qualitative and differential examination of leukocytes. This study proposes a systematic approach to the problem of leukocyte classification in blood smears. However, federated learning techniques cause a lot of communication overhead because of the large weights transmitted and received from the client-side trained models. The purpose of this research is to provide a solution to this issue by combining the strengths of the Whale Optimisation algorithm with Federated Learning (FL). The FL framework that is aided by Whale Optimization Algorithm (WOA) is tested using the Leukocytes images for Segmentation and Classification (LISC) dataset. When compared to the existing federated average model, the suggested approach outperformed it in terms of network efficiency, data imbalance scenarios and communication costs. The suggested framework achieved a prediction accuracy of 97.3% after validation. The results from the suggested system can be a valuable part of leukocyte classification.