Computational Pathology Before and After the Foundation Model Era: Yang Hu, 02/02/26

TIA Centre Seminar Series: Dr Yang Hu Full Title: Computational Pathology Before and After the Foundation Model Era Abstract: Computational pathology is broadly utilized in both diagnosis and biomedical research, offering data-driven approaches to augment traditional histopathological practice. Unlike conventional image analysis, whole slide images (WSIs) present unique challenges due to their extreme size, with weakly supervised learning at the slide level becoming a dominant paradigm. The interplay of multi-scale features, ranging from cellular to tissue-level structures, introduces subtle yet profound influences on the understanding of tissue morphology. In this talk, I will begin by discussing cross-scale feature communication, and then turn to the interpretability of patch-level representations in the era of pathology foundation models. Building on this, I will explore how diverse morphological explanations arise, and conclude with perspectives on the integration and coordination of multiple foundation models to advance computational pathology.