Article Impact Level: HIGH Data Quality: STRONG Summary of Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-28829-5_24 Dr. Gaurang Sharma et al.
Points
- International researchers developed an end-to-end system architecture to securely manage cardiovascular risk assessment and patient clinical monitoring within data-intensive health environments.
- The system uses decentralized federated learning infrastructure to train predictive models across separate storage locations without transferring or centralizing sensitive patient files.
- Validation testing confirmed that artificial intelligence models trained using this privacy-preserving method performed as effectively as traditional centralized machine learning models.
- For diagnosed patients requiring continuous secondary prevention the architecture provides privacy-aware workflows for electronic health record integration and electrocardiogram data collection.
- The unified platform feeds data into secure clinical decision-support tools that adapt to patient records while mitigating automated data hallucinations during clinical reviews.
Summary
This study evaluated the feasibility of implementing a privacy-first, end-to-end system architecture to manage cardiovascular disease (CVD) risk assessment and clinical monitoring. Deploying artificial intelligence (AI) in health networks is routinely challenged by scattered, heterogeneous systems containing highly sensitive patient records that require rigorous privacy protection. To address these data-intensive barriers, researchers from the international Secur-e-Health project engineered a decentralized framework that unifies secure server deployment, granular patient consent management, electronic health record (EHR) integration, and privacy-preserving analytics. The research sought to determine if an integrated digital architecture could support clinical workflows across both primary and secondary prevention pathways without centralizing raw medical files or compromising identity protection.
For primary prevention, the system utilized real-world federated learning (FL) infrastructure to build predictive risk models across multiple distinct data silos. This approach trained machine learning algorithms locally at individual storage sites, completely eliminating the need to transfer or centralize sensitive patient records. Validation testing demonstrated that predictive models developed via this privacy-preserving federated architecture achieved diagnostic performance parity with models trained using traditional, centralized machine learning methods. For secondary prevention, the architecture introduced privacy-aware workflows to capture ongoing electrocardiogram (ECG) monitoring streams, process dynamic record updates, and manage digital treatment consent. These aggregated streams fed into secure clinical decision-support tools designed to minimize automated hallucinations during diagnostic review.
The findings demonstrate that decentralized AI tools can be successfully embedded into the entire cardiovascular care pathway while maintaining absolute institutional data control. By synthesizing secure data processing, careful consent practices, and non-centralized training, the architecture provides a functional framework for multi-institutional collaborative research. While future clinical implementation trials are required to determine specific hazard ratios for long-term adverse cardiac events, this pilot demonstrates a generalizable strategy to optimize digital health services. This system effectively resolves the core tension between clinical data utility and patient confidentiality, offering a scalable blueprint that is highly applicable to other data-intensive medical specialties.
Link to the article: https://link.springer.com/chapter/10.1007/978-3-032-28829-5_24
References
Sharma, G., Pajula, J., Granlund, T., Alhainen, P., Kiljander, T., Bardhi, O., Lähteenmäki, J., Illikainen, A., Väänänen, A., Lipsonen, N., Kaari, S., Salaspuro, V., & Hilvo, M. (2026). End-to-end architecture for secure cardiovascular disease risk assessment and clinical care. In M. Särestöniemi, D. Singh, E. Jarva, & J. Reponen (Eds.), Digital Health and Wireless Solutions: Connected Digital Health: Digital Twins, Wearables, Wireless Systems, and Secure Architectures (pp. 333–355). Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-28829-5_24
