报告主题：Trustworthy AI in Healthcare: From Theory to Medicine Application

报告人：Ling Huang

报告时间： 2026年7月16日 下午4点

报告地点：计算机楼A411

报告摘要：

Building on Dempster-Shafer theory and the random fuzzy set framework, this talk turns to two real-world prediction tasks in medicine where reliable, explainable uncertainty is arguably even more critical to clinical trust. The first part briefly introduces deep evidential fusion for multimodal medical image segmentation, showing how uncertainty quantification combined with a learned reliability coefficient lets a model fuse multiple imaging modalities while flagging low-confidence regions rather than committing to an overconfident boundary. The second part is a line of work grounded in evidential and random fuzzy set theory for time-to-event prediction. This talk starts with calibrated uncertainty quantification, in which relative likelihoods and belief functions are propagated through survival models to produce a full predictive distribution — a possible event interval together with its aleatory and epistemic uncertainty — rather than a single event point estimate. It then extends this to multimodal survival analysis via fusion strategies that combine heterogeneous sources while remaining robust to missing modalities, a common failure mode in real clinical data. Two architectures illustrate this: a dual-prototype evidential fusion model for interpretable WSI survival prediction, and a semantic-anchored fusion approach that incorporates VQA-driven semantic information to improve robustness across imaging centers and clinical domains. Across both tasks, the common thread is that explainability and uncertainty are not features bolted onto a prediction after the fact; they arise jointly from the same evidential representation, providing not just a prediction, but a transparent account of how confident it is and why. 

报告人简介：

Ling Huang received her Master from Zhejiang University of Technology, China in 2019 and her Ph.D. in 2023 from the National Center for Scientific Research (CNRS) lab at Université de Technologie de Compiègne, France. She is currently a postdoctoral researcher at Imperial College London, UK, and a visiting research fellow at the National University of Singapore. Her research interests include the theory of belief functions, with applications to uncertainty quantification, information fusion, causal inference, and decision explanation in medicine. She has published more than 40 papers, including over 20 in top-tier journals and conferences such as Information Fusion, IEEE Transactions on Fuzzy Systems, IEEE Transactions on Cybernetics, Medical Image Analysis, Int. Journal of Approximate Reasoning, ICML, ECCV, ACL, ACM MM, and MICCAI. She serves as an editorial board member of Information Fusion.