A Robust Framework for Secure Cardiovascular Risk Prediction: An Architectural Case Study of Differentially Private Federated Learning

The paper introduces FedCVR, a robust Federated Learning framework designed for secure cardiovascular risk prediction across heterogeneous clinical networks, emphasizing its application of Differential Privacy (DP). It demonstrates that integrating server-side momentum as a temporal denoiser enables the model to achieve an F1 score of 0.78 and an AUC of 0.96 while maintaining a privacy budget (epsilon ~ 13.4). This research highlights the importance of server-side adaptivity in recovering clinical utility under privacy constraints, offering a validated framework for multi-institutional collaboration in AI-driven healthcare.