FEDERATED LEARNING FOR PRIVACY-PRESERVING HEALTHCARE DATA SHARING: ENABLING GLOBAL AI COLLABORATION

Abstract

This study provides a comprehensive systematic review of federated learning as a framework for privacy-preserving healthcare data sharing and its potential to enable global artificial intelligence collaboration. In total, 124 peer-reviewed articles were examined following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure transparency, rigor, and reproducibility. The review highlights how federated learning has evolved from conceptual discussions to practical applications across multiple healthcare domains, including medical imaging, electronic health records, biosignals, and genomic analysis. Key findings indicate that federated architectures, particularly server–client models, have become the dominant deployment strategy, while peer-to-peer approaches are gaining attention for their resilience and decentralization. Privacy-preserving mechanisms—such as differential privacy, secure aggregation, and cryptographic computation—emerged as central to ensuring compliance with regulatory and ethical standards, with adaptive strategies allowing for an effective balance between confidentiality and model utility. Evidence from multi-institutional collaborations shows that federated learning not only improves predictive performance but also enhances inclusivity, enabling smaller or resource-limited institutions to contribute meaningfully without relinquishing data ownership. At the same time, empirical studies identified adversarial risks such as gradient inversion, membership inference, and poisoning attacks, underscoring the necessity for layered safeguards and strong governance structures. Collectively, the findings demonstrate that federated learning is more than a technical innovation; it represents a socio-technical paradigm that integrates privacy, equity, and collaboration into the development of global healthcare AI. This review positions federated learning as a cornerstone for building secure, ethical, and scalable artificial intelligence systems that address the dual imperatives of advancing medical innovation while safeguarding patient confidentiality.