Medical AI Foundation Models
Medical AI foundation models are large-scale, pre-trained models used in healthcare, diagnostics, and biomedical research. Trained on diverse datasets, these models can integrate and reason over multimodal medical data, such as imaging, clinical text, and physiological signals, thereby enabling flexible adaptation across a wide range of clinical tasks. Within the paradigm of physical AI, these models serve as a unifying intelligence layer for systems operating in real-world healthcare environments, with embodied AI systems, including healthcare robots, being a key example. They enable context-aware perception, reasoning over patient-specific data, and execution of actions guided by high-level natural language instructions.
Key research topics, in collaboration with the German Research Center for Artificial Intelligence (DFKI), include interactive machine learning, explainability (XAI), transparency, fairness, and robustness.
