Pathology Foundation Models for Precision Oncology

PhD Qualifying Examination


Title: "Pathology Foundation Models for Precision Oncology"

by

Mr. Jiabo MA


Abstract:

The digitization of histopathology has precipitated a fundamental shift in
oncology, transitioning the field from subjective qualitative assessment to
quantitative computational analysis. This survey comprehensively examines the
emergence of Foundation Models (FMs) in computational pathology, a paradigm
characterized by self-supervised pre-training on large-scale unlabeled Whole
Slide Images (WSIs). We delineate the historical trajectory from traditional
Convolutional Neural Networks (CNNs), which are fundamentally constrained by
the "annotation bottleneck" to modern Vision Transformers (ViTs) that leverage
gigapixel-scale context to learn robust tissue representations. The report
provides an exhaustive analysis of state-of-the-art architectures, including
Virchow, UNI, and CONCH, detailing their construction, pre-training objectives,
and adaptation strategies for downstream tasks. We further scrutinize the
application of these models in precision oncology, specifically in subtyping
rare tumors, stratifying survival risk, predicting therapeutic response through
the discovery of virtual biomarkers, etc. Finally, we critically discuss the
challenges of deployment, including inference latency, interpretability, and
algorithmic bias, offering a roadmap for the future of generalist medical
artificial intelligence.


Date:                   Monday, 9 February 2026

Time:                   3:00pm - 5:00pm

Venue:                  Room 2132C
                        Lift 22

Committee Members:      Dr. Hao Chen (Supervisor)
                        Dr. Dan Xu (Chairperson)
                        Dr. Terence Tsz Wai Wong (CBE)