Table 3.

Quantitative performance comparison among all ablated models for the ablation study

		Single-modal w/o attention	Multi-modal w/o attention	Single-modal with attention	Multi-modal with attention
Patch level	Accuracy	0.665	0.700	0.800	0.893
	AUC	0.644	0.710	0.820	0.961
	Precision	0.679	0.722	0.802	0.897
	Recall	0.894	0.854	0.908	0.939
Region level	Accuracy	0.673	0.702	0.815	0.940
Patient level	Accuracy	0.600	0.733	0.800	0.933