Table 6.

Quantitative evaluation of different models trained with $Datase{t}_3$ for segmentation of multi-class COVID-19 pneumonia lesions. The best results are highlighted in bold.

Method	ground-glass opacity (GGO)				Consolidation				Average
	DSC	${\text{HD}}_{95}$	MAE	NSD	DSC	${\text{HD}}_{95}$	MAE	NSD	DSC	${\text{HD}}_{95}$	MAE	NSD
U-Net	0.3596	7.3888	0.0320	0.3391	0.5277	3.6676	0.0030	0.4838	0.4437	5.5282	0.0175	0.4115
nnU-Net	0.4049	7.7792	0.0214	0.3395	0.4239	4.5909	0.0051	0.3697	0.4144	6.1851	0.0133	0.3546
Inf-Net	0.3021	8.9342	0.0448	0.3084	0.3987	4.8367	0.0054	0.2934	0.3504	6.8855	0.0251	0.3009
SSA-Net	0.4152	6.7788	0.0186	0.3713	0.4953	3.5529	0.0029	0.4529	0.4553	5.1659	0.0108	0.4121
SSA-Net(I)	0.4654	5.9266	0.0116	0.4016	0.6562	2.9541	0.0026	0.6239	0.5608	4.4404	0.0071	0.5128