Universidad Politécnica de Madrid develops privacy-preserving clinical AI

Researchers at the Universidad Politécnica de Madrid (UPM) have developed a new artificial intelligence system called FedSDS designed to improve clinical predictions while strictly maintaining patient privacy.

FedSDS utilizes federated learning and synthetic data to predict the timing of significant clinical events, such as relapses, complications, or patient mortality. By using locally generated synthetic information that reproduces statistical patterns without exposing real medical histories, the system allows hospitals to train robust predictive models without the need to centralize or exchange sensitive patient data.

This approach addresses a critical challenge in healthcare where hospitals are often unable to share real clinical data due to privacy regulations. The technology is noted as being particularly valuable for studying rare diseases, where datasets are typically small, incomplete, or highly varied across different institutions.