Kai Standvoss, Miriam Hägele, Rosemarie Krupar, Julika Ribbat-Idel, Jennifer Altschüler, Gerrit Erdmann, Hans Pinckaers, Evelyn Ramberger et al.
Atlas H&E-TME is an AI-based tissue profiling system that predicts tissue quality, region, and cell types from H&E-stained pathology images, providing over 4,500 quantitative readouts per slide.
H&E staining is the standard in pathology, but quantitative and scalable analysis is challenging. Existing AI systems struggle to achieve expert-level accuracy due to morphological ambiguity and annotation scalability issues.
Leveraging the Atlas family of pathology foundation models, the system analyzes H&E whole-slide images. A multi-pathologist consensus protocol incorporating IHC information improves annotation reliability, and the system is extensively validated on over 1,500 cases across eight cancer types.
Atlas H&E-TME matches or exceeds pathologist performance on H&E alone against an IHC-informed consensus benchmark, demonstrating consistent generalization across diverse cancer types and scanner models. It transforms H&E slides into a scalable quantitative window, laying the foundation for next-generation tissue-based biomarker research.