Table 3.

Model comparison for GGO segmentation.

Methods	Pre-trained architecture	GGO segmentation						Consolidation segmentation
		DSC$\uparrow$	Sens$\uparrow$	Spec$\uparrow$	${\mathit{S}}_{\mathit{\alpha }}\uparrow$	${\mathit{E}}_{\mathit{\varphi }}\uparrow$	MAE$\downarrow$	DSC$\uparrow$	Sens$\uparrow$	Spec$\uparrow$	${\mathit{S}}_{\mathit{\alpha }}\uparrow$	${\mathit{E}}_{\mathit{\varphi }}\uparrow$	MAE$\downarrow$
FCN8s [7]	VGG16	0.482	0.552	0.917	0.591	0.788	0.098	0.289	0.281	0.728	0.573	0.581	0.058
DeepLabV3+ (s$=$8) [14]	ResNet101	0.402	0.501	0.871	0.553	0.682	0.121	0.157	0.173	0.744	0.511	0.556	0.065
DeepLabV3+ (s$=$16) [14]	ResNet101	0.457	0.728	0.845	0.559	0.673	0.149	0.245	0.322	0.721	0.526	0.619	0.079
U-Net [9]	VGG16	0.462	0.374	0.988	0.564	0.731	0.079	0.421	0.427	0.978	0.581	0.781	0.053
SE-Net [18]	–	0.508	0.415	0.889	0.541	0.751	0.075	0.449	0.467	0.958	0.554	0.797	0.051
Semi-Inf-Net-FCN8s [39]	Res2Net + VGG16	0.657	0.731	0.954	0.722	0.884	0.073	0.318	0.251	0.819	0.582	0.588	0.043
Semi-Inf-Net & MC [39]	VGG16 + Res2Net	0.639	0.631	0.973	0.715	0.904	0.070	0.471	0.527	0.979	0.618	0.781	0.045
*FSS-2019-nCov	Res2Net	0.679	0.768	0.980	0.739	0.894	0.061	0.529	0.534	0.983	0.661	0.797	0.045

$\uparrow$denote ‘higher is better’, $\downarrow$denote ‘lower is better’.