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

Disciplines

Other Computer Engineering

Abstract (300 words maximum)

Cardiovascular diseases represent a significant global health concern, accounting for 31% of all worldwide deaths. While machine learning presents a promising avenue for early and accurate diagnosis, the associated ethical and legal challenges, especially concerning data privacy, complicate its direct application. This research paper delves into Federated Learning (FL), a decentralized method, as a potential solution to address data utility and privacy concerns. FL enables devices or servers to hold subsets of overall data, compute local updates, and relay them to a central server without transferring raw data, thus maintaining privacy. The study aims to evaluate the feasibility and efficacy of applying FL to heart disease prediction while maintaining ethical and legal standards. Prior work in this domain, particularly by Wanyong et al., utilized FL for heart sound analysis, highlighting its advantages in data privacy and decentralization. Drawing on this background, our research contributes to the dual objectives of enhancing healthcare outcomes and ensuring data privacy, setting a benchmark for the future application of machine learning in medical research.

Academic department under which the project should be listed

CCSE - Information Technology

Primary Investigator (PI) Name

Seyedamin Pouriyeh

This document is currently not available here.

Share

COinS
 

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

Cardiovascular diseases represent a significant global health concern, accounting for 31% of all worldwide deaths. While machine learning presents a promising avenue for early and accurate diagnosis, the associated ethical and legal challenges, especially concerning data privacy, complicate its direct application. This research paper delves into Federated Learning (FL), a decentralized method, as a potential solution to address data utility and privacy concerns. FL enables devices or servers to hold subsets of overall data, compute local updates, and relay them to a central server without transferring raw data, thus maintaining privacy. The study aims to evaluate the feasibility and efficacy of applying FL to heart disease prediction while maintaining ethical and legal standards. Prior work in this domain, particularly by Wanyong et al., utilized FL for heart sound analysis, highlighting its advantages in data privacy and decentralization. Drawing on this background, our research contributes to the dual objectives of enhancing healthcare outcomes and ensuring data privacy, setting a benchmark for the future application of machine learning in medical research.