JMIR Publications

ABSTRACT

Background:

The exponential growth of medical data and advancements in artificial intelligence have accelerated the development of data-driven healthcare. However, the secure and efficient sharing of sensitive medical data across institutions remains a major challenge due to privacy concerns, data silos, and regulatory restrictions. Traditional centralized systems are prone to data breaches and single points of failure, while existing privacy-preserving techniques face high computational and communication costs.

Objective:

This study aims to provide a comprehensive review of the recent advances in Blockchain-based Federated Learning (BCFL) within the medical field. By exploring the synergistic integration of federated learning and blockchain, this review evaluates how BCFL enhances data security, supports privacy-preserving cross-institutional collaboration, and facilitates practical applications in healthcare, including medical data sharing, IoMT, public health surveillance, and telemedicine.

Methods:

We conducted a systematic literature review using databases such as PubMed, IEEE Xplore, Web of Science, and Google Scholar. Boolean logic and domain-specific keywords were employed to retrieve studies from 2018 to 2024. After automated deduplication and multi-stage manual screening, over 100 high-quality papers were included. These works cover BCFL’s theoretical foundations, system architectures, application domains, limitations, and future directions.

Results:

BCFL frameworks combine the decentralized trust and auditability of blockchain with the privacy-preserving collaborative learning capabilities of FL. This integration mitigates risks such as model tampering, data leakage, and lack of incentives in federated systems. Applications span across cross-institutional medical data sharing, Internet of Medical Things (IoMT), epidemic forecasting, and telemedicine. Architectures including fully coupled, flexibly coupled, and loosely coupled models offer varying trade-offs between efficiency, scalability, and security.

Conclusions:

Blockchain-based federated learning represents a transformative paradigm for secure, collaborative, and privacy-preserving medical AI. By combining decentralized trust, incentive-driven participation, and privacy-enhancing machine learning, BCFL paves the way for next-generation smart healthcare systems. Despite current technical and practical challenges, BCFL demonstrates strong potential to support precision medicine, global health data collaboration, and large-scale AI deployment in healthcare.