Improving Performance Metrics on FedAvg with Blockchain for Privacy-Preserving Healthcare AI

The demand for AI and its applications in healthcare provides significant opportunities for improving disease prediction, disease identification, clinical decision-making, and healthcare treatment. In healthcare, patient privacy-preserving, data security, and confidentiality continue to restrict the sharing of medical data across healthcare institutions. To address these challenges and improve the performance metrics, this study proposed a secure and privacy-preserving healthcare model that integrates the Federated Averaging (FedAvg) algorithm with Blockchain technology. The proposed FedAvg enables multiple healthcare institutions to collaboratively train AI models while retaining sensitive patient data using the FedAvg algorithm with Blockchain privacy-preserving healthcare AI. Rather than transferring raw data, only encrypted model parameters are exchanged using healthcare AI and aggregated to construct a global learning model using the FedAvg algorithm. A permissioned blockchain concept is incorporated to ensure trust, transparency, and immutability throughout the FL process. Smart contracts ensure to validate model updates, while a Practical Byzantine Fault Tolerance (PBFT) consensus mechanism facilitates secure agreement among participating entities for transferring data between healthcare institutions. Secure FedAvg with Blockchain for privacy-preserving healthcare AI was evaluated on publicly available healthcare datasets. This article used a heart-disease dataset and assessed it using performance metrics such as accuracy, latency, communication overhead, and scalability. Experimental results proved that the proposed architecture achieved high prediction accuracy while increasing the number of rounds and model accuracy also increased.