Semi-supervised COVID-19 CT image segmentation using deep generative models

BMC Bioinformatics

Table 1 Quantitative results of ground-glass opacity (GGO), consolidation (CON), background, and the overall average on the test dataset

Lesion	Method	IoU	F1	Recall	Precision
GGO	U-Net	0.391 ± 0.280	0.499 ± 0.32	0.608 ± 0.358	0.47 ± 0.326
GGO	SegNet	0.004 ± 0.027	0.007 ± 0.044	0.012 ± 0.087	0.009 ± 0.071
GGO	StitchNet	0.358 ± 0.257	0.471 ± 0.303	0.517 ± 0.331	0.489 ± 0.328
CON	U-Net	0.404 ± 0.331	0.49 ± 0.368	0.616 ± 0.378	0.485 ± 0.38
CON	SegNet	0.021 ± 0.113	0.027 ± 0.137	0.057 ± 0.227	0.021 ± 0.114
CON	StitchNet	0.318 ± 0.315	0.397 ± 0.361	0.539 ± 0.411	0.387 ± 0.369
Background	U-Net	0.983 ± 0.023	0.992 ± 0.012	0.987 ± 0.02	0.996 ± 0.006
Background	SegNet	0.97 ± 0.044	0.984 ± 0.024	0.999 ± 0.009	0.971 ± 0.043
Background	StitchNet	0.985 ± 0.021	0.992 ± 0.011	0.992 ± 0.011	0.993 ± 0.014
Overall	U-Net	0.593	0.66	0.737	0.65
Overall	SegNet	0.332	0.339	0.356	0.334
Overall	StitchNet	0.554	0.62	0.683	0.623

ISSN: 1471-2105