GR-406 Federated Learning in Cardiac Diagnostics: Balancing Predictive Accuracy with Data Privacy in Heart Sound Classification

GR-406 Federated Learning in Cardiac Diagnostics: Balancing Predictive Accuracy with Data Privacy in Heart Sound Classification

https://ccse.kennesaw.edu/computing-showcase/cday-programs/fall23program.php

Cardiovascular diseases account for nearly a third of global deaths, posing a challenge that machine learning can help address. However, data privacy concerns hinder the direct application of conventional machine learning in this sensitive area. This paper explores Federated Learning (FL) as a decentralized strategy to mitigate these concerns by allowing for local data processing. FL's design ensures that only processed updates, not raw data, are shared with a central server, maintaining individual privacy. Our research assesses FL's practicality and effectiveness in predicting heart disease while adhering to ethical and legal norms. We build upon previous studies, such as Wanyong et al.'s work on heart sound analysis with FL, to underline its privacy-preserving benefits. This study aims to improve healthcare outcomes with machine learning while setting a privacy-conscious benchmark for future research.