Deep learning predicts pancreatic cancer molecular subtypes from routine histopathology

This study explores the use of deep learning to infer molecular subtypes of pancreatic ductal adenocarcinoma directly from routine hematoxylin and eosin–stained histology slides. By analyzing data from 689 patients with paired histology and RNA sequencing, the researchers developed PanSubNet, a model that integrates cellular morphology and tissue architecture to predict cancer subtypes, potentially overcoming limitations of traditional transcriptomic subtyping such as cost and tissue requirements.

Originally published by gnews