Federated deep learning for privacy-preserving cardiovascular disease risk prediction
5 affiliations
- 1Department of Radiology and Nuclear Medicine, Erasmus MC University Medical Center Rotterdam, Rotterdam, Netherlands
- 2Netherlands eScience Center, Matrix THREE, Amsterdam, Netherlands
- 3Department of Epidemiology, Erasmus MC University Medical Center Rotterdam, Rotterdam, Netherlands
- 4Department of Cardiology, Erasmus MC University Medical Center Rotterdam, Rotterdam, Netherlands
- 5Department of Cardiology, University Medical Center Utrecht, Utrecht, Netherlands
Abstract
Cardiovascular disease risk prediction models often rely on data from a single institution or centrally pooled datasets. Extending these models across institutions could be limited by privacy regulations and constraints on sharing patient-level data. Federated learning enables collaborative model development without transferring sensitive patient data, but its application in healthcare remains challenging because datasets often differ in size, population characteristics, and outcome definitions. In this study, we present a federated deep learning approach for privacy-preserving cardiovascular disease risk prediction that integrates two population-based cohorts with different characteristics: Lifelines, including 148,230 participants meeting the study inclusion criteria with self-reported outcomes, and the Rotterdam Study, including a smaller cohort of 10,155 participants with digitally linked clinical outcomes. Model performance was primarily evaluated on the Rotterdam Study because of its complete follow-up. Deep survival models trained using federated learning achieved a higher mean C-statistic than models trained locally without federation. For the Rotterdam Study, the C-statistic increased from 0.723 (95% CI: 0.714–0.732) to 0.737 (95% CI: 0.727–0.747). For Lifelines, the C-statistic increased from 0.791 (95% CI: 0.783–0.799) to 0.795 (95% CI: 0.785–0.804). These findings suggest that federated deep learning across heterogeneous cohorts can improve the discrimination of cardiovascular disease risk prediction models while keeping individual-level patient data within the participating institutions.
Related
- PA-CFL: Privacy-Preserving Adaptive Clustered Federated Learning Under Heterogeneous Data
- Privacy-Preserving Over-the-Air Federated Learning over Cell-Free Massive MIMO System with User-Centric Clustering
- FedCap-Priv: Geometry-Aware Privacy and Bandwidth-Efficient Routing for Federated Capsule Learning
- Projection-Based Federated Learning on Compact Submanifolds with Geometry-Independent Differential Privacy
- Privacy Preserving Gossip Learning
BibTeX
@article{mo2026federated,
title = {{Federated deep learning for privacy-preserving cardiovascular disease risk prediction}},
author = {Hyunho Mo and Djura Smits and Mahlet A. Birhanu and Maarten J. G. Leening and Maarten J. G. Leening and Maarten J. G. Leening and Daniel Bos and Daniel Bos and Pim van der Harst and Esther E. Bron},
journal = {Frontiers in Applied Mathematics and Statistics},
year = {2026},
month = oct,
doi = {10.3389/fams.2026.1923866},
url = {https://doi.org/10.3389/fams.2026.1923866},
note = {CC BY},
}
RIS
TY - JOUR
TI - Federated deep learning for privacy-preserving cardiovascular disease risk prediction
AU - Hyunho Mo
AU - Djura Smits
AU - Mahlet A. Birhanu
AU - Maarten J. G. Leening
AU - Maarten J. G. Leening
AU - Maarten J. G. Leening
AU - Daniel Bos
AU - Daniel Bos
AU - Pim van der Harst
AU - Esther E. Bron
PY - 2026
DA - 2026/10/23
JO - Frontiers in Applied Mathematics and Statistics
DO - 10.3389/fams.2026.1923866
UR - https://doi.org/10.3389/fams.2026.1923866
AB - Cardiovascular disease risk prediction models often rely on data from a single institution or centrally pooled datasets. Extending these models across institutions could be limited by privacy regulations and constraints on sharing patient-level data. Federated learning enables collaborative model development without transferring sensitive patient data, but its application in healthcare remains challenging because datasets often differ in size, population characteristics, and outcome definitions. In this study, we present a federated deep learning approach for privacy-preserving cardiovascular disease risk prediction that integrates two population-based cohorts with different characteristics: Lifelines, including 148,230 participants meeting the study inclusion criteria with self-reported outcomes, and the Rotterdam Study, including a smaller cohort of 10,155 participants with digitally linked clinical outcomes. Model performance was primarily evaluated on the Rotterdam Study because of its complete follow-up. Deep survival models trained using federated learning achieved a higher mean C-statistic than models trained locally without federation. For the Rotterdam Study, the C-statistic increased from 0.723 (95% CI: 0.714–0.732) to 0.737 (95% CI: 0.727–0.747). For Lifelines, the C-statistic increased from 0.791 (95% CI: 0.783–0.799) to 0.795 (95% CI: 0.785–0.804). These findings suggest that federated deep learning across heterogeneous cohorts can improve the discrimination of cardiovascular disease risk prediction models while keeping individual-level patient data within the participating institutions.
SN - 2297-4687
ER -