Personalizing Federated Learning Models: Methods, Applications, and Challenges in Healthcare 5.0

Personalized Federated Learning (PFL) revolutionizes healthcare by enabling collaborative model training across decentralized institutions while preserving patient confidentiality. The integration of Foundation Models (FMs) such as Generative Pretrained Transformer (GPT), Vision Transformer (ViT), and Contrastive Language Image Pretraining (CLIP) further enhances personalization in federated settings by addressing challenges such as data heterogeneity and local resource limitations. This paper comprehensively reviews FM driven PFL, including its systematic classification, mathematical foundations, and applications in medical imaging. It also outlines key research challenges and proposes future directions to advance PFL within the framework of Healthcare 5.0. The synergy between PFL and FMs enables the development of more robust and universal AI solutions suitable for diverse clinical environments. Moreover, multimodal personalization paves the way for precision diagnosis by integrating imaging, genomics, and clinical narratives. These innovations are foundation to building secure, equitable, and intelligent healthcare systems.