Advancing Privacy Preservation in Healthcare Through Federated Learning

Abstract

Artificial Intelligence and machine learning are now applied to healthcare, which is changing the possibility of new ways of diagnosing diseases, additional diagnostic methods, and individualized treatments. However, due to the large volumes of sensitive patient information, there are still many barriers constraining the creation of reliable and transferable bio-artificial intelligence models across institutions. Federated learning has risen as a unique solution effectively training models together but not sharing raw data which is the core concern of centralized learning. Federated learning is the primary research subject of this paper with an emphasis on the opportunity to maintain patient privacy whilst improving the accuracy and relevance of predictive models in healthcare. A discussion of previous work is provided to identify significant contributions and discuss the limitations of studies, including data heterogeneity, privacy, and regulations. To address these issues, a novel FL framework for healthcare is proposed below. The study ends with suggestions pointing to how to take FL a notch higher in healthcare and some of them are the issues of flexibility which remains a significant barrier to the integration of Federated learning.

Ähnliche Arbeiten