Detection of Heart Disease Using Deep Federated Learning Algorithms

Abstract

Heart disease remains one of the leading causes of mortality worldwide, necessitating advanced and efficient diagnostic methods. Traditional machine learning (ML) as well as deep learning (DL) approaches for heart disease detection often rely on centralized data, which raises concerns about data privacy, security, and accessibility. Recently, Federated Learning (FL), a decentralized machine learning paradigm, has emerged as a promising solution to these challenges. This paper explores the application of Deep Federated Learning (DFL) algorithms in the detection of heart disease namely Federated Averaging (FedAvg), Federated Learning with Differential Privacy (DP-FL), Federated Transfer Learning (FTL), Personalized FL (PFL) and Lightweight FL (LFL).We also discuss the principles of FL, its integration with deep learning models, and its advantages in healthcare specially for heart diseases detection. Additionally, we review recent advancements, challenges, and future directions in this field. Finally, it is observed that Personalized Federated Learning (PFL) achieves the highest performance with 96.2% accuracy, 96.4% F1-score that demonstrating the benefits of adapting models to individual client data.