Trust and Accountability in Federated Healthcare AI: A Blockchain-Governed Systematic Review

Abstract

Although federated learning finds wide application in healthcare, the development of effective governance frameworks that incorporate strong privacy protections and regulatory compliance remains behind schedule, especially for sensitive areas like mental health. In this systematic review, we analyze blockchain-enabled governance in healthcare federated learning, paying particular attention to consent management, auditability, privacy budget monitoring, and regulatory compliance. Following the PRISMA 2020 methodology, we performed a search of seven bibliographic databases for peer-reviewed literature published between January 2020 and March 2026, identifying 3,847 sources, of which 67 papers were included. We find significant heterogeneity in governance solutions: on-chain privacy budget tracking is available in only 23%, dynamic consent with revocation in 31%, and security analysis in 42%. Ethereum dominates the implementation landscape, and mental health applications are underutilized despite increased privacy considerations. Performance overheads are large, with the median latency increase of 340% compared to conventional federated learning. Moreover, we provide an industry-ready implementation of an Ethereum-based governance framework for mental health depression detection, providing efficient consent validation and audit trail generation, and protecting user privacy without sacrificing model accuracy.