Trustworthy Clinical AI
Reliable, explainable, and human-centered medical AI systems.
This direction focuses on medical AI that clinicians can inspect and use with confidence. Current interests include low-dose reconstruction, transparent anomaly localization, multimodal neuroimaging analysis, and human-in-the-loop decision support.
The central question is not only whether a model is accurate, but whether its evidence, limitations, and behavior are understandable in the clinical workflow.