. 2022 Jun 27;252:109278. doi: 10.1016/j.knosys.2022.109278

Table 1.

Comparison of classification results of different methods on three targets (bag size=100).

Targets	Method	ACC( $%$ )	AUC( $%$ )
COVID-19 $v s .$ NC	ResNet18-Voting	88.97 $[85.52, 92.37]$	90.08 $[86.30, 92.18]$
	Gated-Attention	94.48 $[92.07, 97.24]$	94.63 $[91.65, 97.08]$
	EM-Cls	96.73 $[95.00, 98.80]$	96.98 $[93.80, 98.99]$
	EMTN	98.62 $[97.59, 100.0]$	98.90 $[97.65, 100.0]$

COVID-19 $v s .$ CP	ResNet18-Voting	81.28 $[76.68, 85.74]$	81.63 $[78.03, 84.08]$
	Gated-Attention	90.34 $[87.19, 93.85]$	90.68 $[86.95, 93.15]$
	EM-Cls	93.10 $[90.33, 95.89]$	92.53 $[89.83, 95.72]$
	EMTN	95.17 $[92.28, 97.38]$	95.87 $[91.49, 98.65]$

CP $v s .$ NC	ResNet18-Voting	84.93 $[80.86, 89.15]$	85.72 $[81.38, 88.65]$
	Gated-Attention	91.03 $[87.99, 94.43]$	91.00 $[87.19, 94.22]$
	EM-Cls	93.79 $[91.15, 96.45]$	94.03 $[90.39, 97.12]$
	EMTN	95.86 $[93.95, 98.23]$	96.46 $[92.97, 98.79]$

The upper and lower bounds of 95% confidence interval are shown in [ $\cdot$ ].