Inference of tumor spatial habitats

Stanford researchers developed CANVAS, a model that predicts spatial tumor microenvironments from routine H&E histopathology images by learning from large spatial proteomics data of lung cancer patients. CANVAS uses pretrained pathology models and data augmentation to infer cellular neighborhoods, enabling scalable analysis of tumor habitats without costly spatial omics.