A Review on Privacy-Preserving Techniques in Federated Learning for Medical Image Analysis
K. Hemalatha, Shradhanjali Das, Ananya Sundar, A. M. · 2025
Federated learning (FL) is transforming healthcare by enabling collaborative AI model development across institutions while ensuring data privacy. In medical imaging, FL facilitates the use of decentralized datasets without compromising patient confidentiality, addressing privacy regulations crucial to clinical settings. This survey examines how FL can be applied across different radiological modalities, such as X-ray, CT (Computed Tomography), and MRI (Magnetic resonance imaging). It emphasizes the potential for FL to enhance diagnostic accuracy while addressing challenges like data variability and privacy concerns. In this paper, we critically analyze the gap between advanced FL techniques and their practical deployment in health-care. We also review innovative solutions, including differential privacy mechanisms and semi-supervised learning, to mitigate risks like data leakage and model inversion attacks. Additionally, alternative frameworks such as swarm learning and peer-to-peer systems are discussed for enhancing scalability. By synthesizing current research, we identify key challenges and opportunities for optimizing FL in healthcare, proposing future directions to ensure its effective integration into clinical workflows.