Table 2.

WSI level accuracy and AUC of each model.

Model name	Acc	AUC (95% CI)	Sensitivity	Specificity	PPV	NPV	Cohort
densenet121	0.954	0.992 (0.9915–0.9923)	0.952	0.957	0.954	0.955	Train
densenet121	0.586	0.682 (0.6761–0.6878)	0.771	0.530	0.336	0.882	Test
resnet101	0.974	0.997 (0.9968–0.9972)	0.974	0.974	0.972	0.975	Train
resnet101	0.502	0.639 (0.6332–0.6451)	0.843	0.397	0.302	0.891	Test
inception_v3	0.978	0.998 (0.9978–0.9981)	0.981	0.976	0.974	0.982	Train
inception_v3	0.542	0.612 (0.6057–0.6183)	0.681	0.499	0.296	0.835	Test