Fig. 4: Per-slide AUC values for cross-classification of tumor/normal status. | Nature Communications

Fig. 4: Per-slide AUC values for cross-classification of tumor/normal status.

From: Deep learning-based cross-classifications reveal conserved spatial behaviors within tumor histological images

Fig. 4

The hierarchically clustered heatmap shows pairwise AUC values of CNNs trained on the tumor/normal status of one cancer type (train axis) tested on the tumor/normal status of another cancer type (test axis). Adeno-ness (adenocarcinoma vs. non-adenocarcinoma) and organ of origin (lung, kidney, gastrointestinal, gynecological) for each set are marked with colors on the margins. Cancers with ambiguous or mixed phenotype are marked as “Other”.

Back to article page