TWIN-X

TWIN-X will build an interactive digital patient twin for oncology and cardiology that integrates imaging, clinical narratives, laboratory data and pathology to support precision medicine. The twin is a modular, clinically coherent representation that clinical researchers can interrogate and update through transparent queries, what if simulations and counterfactual exploration. Users can compare therapeutic options and timing, inspect predicted trajectories with calibrated uncertainty, trace outputs to underlying evidence and receive concise, verifiable rationales generated from structured sources. Generative AI first structures data before embedding to preserve clinical meaning and enable faithful explanations. Trust is ensured through end to end calibration, out of distribution detection, selective abstention, human oversight and equity audits with mitigation. Data remain within institutions under GDPR through federated training, pseudonymisation and strict governance. Interoperability follows established standards. We will deliver reference implementations, evaluation protocols and model and data cards, and validate prototypes on retrospective European cohorts at TRL 4-5. TWIN-X will advance robust, explainable generative AI tools for health research and prepare documentation required for future conformity assessment under the EU AI Act and Medical Device Regulation. Expected impact is faster evidence generation and auditable decision support in simulated workflows, with no patient facing deployment during the project.