. 2021 Sep 9;211:106406. doi: 10.1016/j.cmpb.2021.106406

Table 2.

Performance comparison of the various models for identifying COVID-19.

Model	ACC	SEN	SPE	AUC	F1 score
M_{ResNet-50-MIL}	0.959	0.972	0.940	0.955	0.965
M_{ResNet-50-MIL-frozen-layers}	0.888	0.887	0.890	0.922	0.903
M_Xception-MIL	0.938	0.965	0.900	0.936	0.948
M_{ResNet-50-Voting}	0.900	0.957	0.820	0.844	0.918
M_{Xception-Voting}	0.920	0.936	0.900	0.898	0.933
M_{ResNet-50-Montages}	0.730	0.965	0.400	0.819	0.807
M_{Xception-Montages}	0.734	0.760	0.690	0.806	0.772
M_MedicalNet	0.681	0.908	0.360	0.670	0.769
M_{ResNet-50-MIL-max-pooling}	0.896	0.936	0.840	0.936	0.914
M_{ResNet-50-MIL-Noisy-AND-pooling}	0.880	0.929	0.810	0.939	0.900