The rapid growth of AI in healthcare has spurred the need for robust machine learning models capable of handling sensitive data while ensuring compliance with privacy regulations. Traditional centralized learning, where data is aggregated on a central server, faces significant challenges related to data privacy, transfer costs, and scalability. Federated learning presents an alternative by enabling decentralized training across multiple data sources while preserving privacy. This paper explores the strengths and weaknesses of centralized and federated learning approaches within the healthcare context, with a focus on privacy concerns, scalability issues, and the ability to handle heterogeneous data. We analyze real-world applications to illustrate the practical implications of each approach and identify key areas where federated learning could outperform centralized models. Additionally, the paper outlines future research directions to improve scalability, privacy-preserving techniques, and algorithmic developments for healthcare AI, aiming to bridge the gap between theoretical advances and practical deployment.

Authors: Yemi Tunrayo Fatokun, Arome Junior Gabriel, Olufunso Dayo Alowolodu, Adebayo O. Adetunmbi

Published in: International Conference for Internet Technology and Secured Transactions (ICITST-2024)

• Date of Conference: 4-6 November 2024
• DOI: 10.20533/ICITST.2024.0010
• ISBN: 978-1-913572-76-1
• Conference Location: St Anne’s College, Oxford University, UK