Klinikum rechts der Isar, Technical University of Munich

TUM Klinikum, Klinikum rechts der Isar is the university hospital of the Technical University of Munich (TUM), a high-volume centre of maximum care with a research environment for AI, digital medicine and clinical translation. In TWIN-X, TUM Klinikum connects cardiology and oncology pathways with machine learning expertise from the AI Assisted Healthcare working group. It contributes secure compute linked to hospital data platforms and reproducible engineering workflows, operated so that sensitive health data remain within the hospital and are processed in place rather than transferred to external infrastructures. This setting allows TUM Klinikum to bridge routine clinical data, radiology, pathology, longitudinal patient records and advanced AI methods. TUM Klinikum therefore provides both clinical and technical capacity for TWIN-X by anchoring the project in real care pathways while supporting open science, quality assurance, data governance, model evaluation and safe processing of sensitive health data.

Role within TWIN-X

TUM acts as the coordinating institution of TWIN-X and leads WP1 Project management and scientific coordination as well as WP2 Representations from unstructured clinical data. In WP1, TUM establishes the General Assembly and Steering Committee, leads scientific coordination, chairs regular Steering Committee reviews, coordinates contractual and governance issues with EURICE, and oversees the Data Management Plan. In WP2, TUM drives the transformation of clinical narratives, radiology reports, discharge letters, coded events, medications and laboratory values into computable patient representations and longitudinal timelines. TUM leads the modular GenAI structuring pipeline, consolidates text-derived patient embeddings and defines the embedding interfaces needed for downstream fusion. Beyond these lead roles, TUM Klinikum contributes to selected tasks in other work packages together with the consortium partners, including radiology reference segments and model engineering, whole-slide pathology pipeline development, cross-modality calibration and confidence schemas, and the Query Retrieval Fusion backbone. It further supports the oncology and cardiology demonstrators with reproducible analysis workflows, uncertainty propagation and audit-ready evidence traces.

Main contacts

Prof. Dr Keno Bressem

TWIN-X Coordinator

Prof. Dr Lisa C. Adams

Dr Cosmin I. Bercea