Development of an AI-based population health foundation model utilizing health and register data from three European countries

In this project, supported by the Research Council of Finland, we develop advanced artificial intelligence (AI) models to analyze health data from national databases in Europe. These AI systems predict future health events and adapt to various healthcare tasks, aiming at more accurate prediction of progression of diseases like cancer, or estimation of future healthcare costs, useful in public health planning. One strength of the project is the use of representative large datasets from Finland (7 million people) and France (16 million people). With the aid of such comprehensive data, it is possible to train more accurate models for minority groups and individuals suffering from rare diseases. A key goal is protecting privacy, aiming to use AI to create synthetic, anonymous datasets that mimic real data but cannot be linked to any real person, minimizing privacy concerns. We seek to make health data AI models more versatile, and improve privacy safeguards, advancing public health and healthcare innovation.

Machine learning for health (ML4HEALTH) – A nationwide, longitudinal and multimodal health data cohort

ML4HEALTH is a Finnish nationwide health register study pulling together longitudinal health records from ~7 million individuals with a follow-up of approximately 25 years for developing modern AI tools for real world health data. The data is hosted at the secure and audited FIMM Sandbox environment.