Multimodal, Generative, and Agentic AI for Pathology

Explore cutting-edge applications of artificial intelligence in digital pathology through this 47-minute conference talk from the Broad Institute's Models, Inference and Algorithms series. Discover data-efficient methods for weakly-supervised whole slide classification with practical applications in cancer diagnosis, subtyping, identifying origins for cancers of unknown primary, and allograft rejection detection. Learn about interpretable multimodal deep learning approaches that integrate histology and genomic data to discover prognostic markers. Examine the development of both unimodal and multimodal foundation models for pathology that contrast with language and genomics models. Understand the creation of universal multimodal generative co-pilots and chatbots specifically designed for pathological analysis. Delve into 3D computational pathology techniques and explore critical considerations around bias and fairness in computational pathology algorithms. Gain insights into agentic AI workflows that enhance diagnostic pathology and biomedical research capabilities. The presentation covers groundbreaking research published in leading journals including Nature, Nature Medicine, Cell, Cancer Cell, and top-tier computer vision conferences, providing comprehensive coverage of the latest advances in AI-driven pathological analysis and diagnosis.