The abstract is read at the publisher; the summary is JClub's.
CANVAS infers tumor ecological habitats from H&E by learning 10 reproducible cellular neighborhoods (CNs) from an atlas of over 18 million cells in non-small cell lung cancer.
- Why it matters: Spatial architecture of the tumor microenvironment is hard to characterize and translate clinically, limiting how well spatial context can inform prognosis and therapy response.
- What they did: CANVAS uses multimodal alignment and foundation-model-based morphological encoding to predict CN-anchored habitat structures from H&E, trained on 41-plex spatial proteomics across 457 non-small cell lung cancer patients.
- The result: In over 5,000 patients across 9 cancer types, CANVAS supports prognostic modeling, spatial ecotype stratification, and immunotherapy outcome prediction, enabling clinically scalable spatial profiling.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.