Fig. 1.

Outline of the study and development of the classifier. A The cohorts used for the present studies and their regional origin. B Overview of study methodology. C Sankey graph of how TCGA classifications changed when independently evaluated by our in-house expert pathologist (HIG). D KMC for OS comparing pathologist-model unanimity and discordance for intestinal and diffuse type using concatenated validation results for the TCGA cross-validation folds. E ROC curve for the five-fold cross validation on the TCGA cohort using only WSIs in which both TCGA and our pathologist were in unanimity with when performing Laurén