A Decade of Federated Learning Applications in Medical Image Analysis in Shanghai: A Comprehensive Review (2014–2024)

Federated learning (FL) has revolutionized medical image analysis by enabling collaborative model training while maintaining patient data privacy. This paper explores the applications, challenges, and innovations of FL in medical imaging, highlighting its role in diagnostic improvements, data heterogeneity management, and integration with explainable AI (XAI). Notable advancements include open-source frameworks, privacy-preserving techniques, and strategies to address non-IID data challenges. Case studies from Shanghai demonstrate FL's potential in enhancing diagnostic accuracy through multi-institutional collaboration while safeguarding sensitive data. Despite these successes, challenges remain, including computational constraints, ethical considerations, and technical complexities. The paper emphasizes future research directions such as refining FL algorithms for heterogeneous data, improving privacy techniques, and enhancing model generalization. By addressing these challenges, FL can continue to transform medical imaging, fostering secure, efficient, and scalable AI applications in healthcare.