Fig. 3

Performance and representative results across different models on the inflammation dataset. The highest performing model is CDAM with a mean AUROC of 0.866 ± 0.029. The remaining performances are: 0.589 ± 0.035 (CLAM), 0.842 ± 0.036 (DAM), 0.838 ± 0.037 (DGAM), and 0.859 ± 0.027 (CDGAM). CDGAM yields the heatmaps with highest segmentation and contrast quality. Note that CLAM highlights many tiles from glass whereas our CDGAM highlights liver tissues which contributes to the diagnoses.