TABLE 4. Quantitative Comparison of the Proposed Model and Four Widely Used Deep Models for Multi-Class Infection Segmentation. (Mean ± Standard Deviation of DSC, Sensitivity, and Specificity, Best Results are Highlighted in Bold).

Method	Infection type	DSC	Sen.	Spec.	P-value
U-Net [21]	Ground-Glass Opacities	0.6034 ± 0.073	0.72 31 ± 0.011	0.9775 ± 0.010
	Interstitial Infiltrates	0.6423 ± 0.081	0.7361 ± 0.025	0.9756 ± 0.008
	Consolidation	0.7526 ± 0.036	0.8209 ± 0.026	0.9863 ± 0.005
U-Net++ [22]	Ground-Glass Opacities	0.7160 ± 0.052	0.8017 ± 0.014	0.9675 ± 0.004
	Interstitial Infiltrates	0.6971 ± 0.034	0.7829 ± 0.018	0.9847 ± 0.008
	Consolidation	0.8041 ± 0.042	0.8172 ± 0.009	0.9865 ± 0.003
Attention U-Net [24]	Ground-Glass Opacities	0.7226 ± 0.026	0.8038 ± 0.019	0.9665 ± 0.012
	Interstitial Infiltrates	0.7158 ± 0.024	0.7953 ± 0.011	0.9812 ± 0.007
	Consolidation	0.8012 ± 0.041	0.8147 ± 0.015	0.9814 ± 0.008
U-Net+CBAM [41]	Ground-Glass Opacities	0.7037 ± 0.039	0.8172 ± 0.013	0.9675 ± 0.005
	Interstitial Infiltrates	0.6824 ± 0.032	0.7975 ± 0.018	0.9431 ± 0.008
	Consolidation	0.8005 ± 0.052	0.8727 ± 0.021	0.9827 ± 0.004
Our network	Ground-Glass Opacities	0.7422 ± 0.038	0.8593 ± 0.018	0.9742 ± 0.005
	Interstitial Infiltrates	0.7384 ± 0.021	0.8268 ± 0.020	0.9869 ± 0.005
	Consolidation	0.8769 ± 0.015	0.8645 ± 0.017	0.9889 ± 0.007