Knowledge-infused Efficient Learning for Computational Pathology

Abstract

Developments in data sharing and computational resources are reshaping digital pathology research. These advancements are enabling a transition from small-scale to large-scale analysis, from 2D to 3D imaging, and from single to multi-modality approaches. Classical deep learning methods and expert examination are becoming insufficient to fully leverage these trends. In this thesis defense, Mr. Ruining Deng will present his research principles, which include building AI-powered computer vision tools with cross-modal knowledge, advancing clinical research with AI-driven tools, democratizing pathology research, and incorporating foundation models to create pathology datasets. His work emphasizes integrating domain knowledge, large-scale modeling, and multi-modal fusion to enhance digital pathology workflows, ultimately aiming to make these tools accessible and usable in clinical settings.