Table 2.

Evaluation of the proposed methods on YUMC patch dataset.

Method	Accuracy	AUC	F1	Specificity	Sensitivity
MIL Campanella et al. (2019)	0.845	0.852	0.836	0.8	0.903
DeepAttentionMIL Ilse et al. (2018)	0.845	0.859	0.845	0.75	0.968
JointMIL Chikontwe et al. (2020)	0.845	0.837	0.814	0.9	0.774
DA-CMIL (w/o ${\mathcal{A}}_{\theta ,S,I}$)	0.718	0.728	0.714	0.65	0.806
DA-CMIL (w/o ${\mathcal{L}}_F$)	0.873	0.88	0.866	0.825	0.935
DA-CMIL (w/ ${\mathcal{L}}_F$)	0.958	0.955	0.951	0.975	0.935