Pathology Foundation Models for Segmentation and Precision Oncology: Junlin Hou, 29/06/26

TIA Centre Seminar Series: Dr Junlin Hou Full Title: Pathology Foundation Models for Text-prompted Segmentation and Precision Oncology Abstract: Computational pathology is rapidly moving from task-specific image analysis toward foundation models that enable scalable, generalizable disease diagnosis and clinical decision support. In this talk, I will first present our recent work, PathSegmentor, a foundation model that enables flexible, text-prompted segmentation across diverse pathological entities. Beyond segmentation, I will also provide an overview of our broader efforts in pathology foundation models, including image enhancement, whole-slide representation learning, benchmarking, and agentic AI systems. These works aim to advance computational pathology toward more precise, interpretable, and clinically useful AI for precision oncology.