Table 5. Evaluation indices for the proposed semi-supervised learning framework with CNN-based methods on the DME dataset.

Network	Metrics
Full-supervised	DSC(%)↑	Jaccard(%)↑	95HD(voxel)↓	ASD(voxel)↓
VNet [29]	88.23 ± 0.87	78.65 ± 1.24	1.86 ± 0.32	0.82 ± 0.43
UNet [33]	88.04 ± 1.54	78.68 ± 2.47	1.82 ± 0.61	0.62 ± 0.28
ResNet34[34]	89.51 ± 0.78	79.33 ± 1.51	1.41 ± 0.44	0.63 ± 0.32
Semi-supervised (25%)
VNet + CML	88.36 ± 1.41	79.29 ± 1.53	1.74 ± 0.32	0.68 ± 0.29
UNet + CML	88.10 ± 3.77	78.93 ± 5.96	2.33 ± 1.60	0.62 ± 0.40
ResNet34 + CML	89.41 ± 1.14	80.87 ± 1.88	1.47 ± 0.41	0.56 ± 0.31
VNet + CML + GRA	91.04 ± 0.24	81.93 ± 0.24	1.38 ± 0.29	0.56 ± 0.27
UNet + CML + GRA	90.28 ± 0.55	80.31 ± 0.47	1.48 ± 0.56	0.63 ± 0.37
ResNet34 + CML + GRA	91.37 ± 0.24	82.57 ± 0.22	1.35 ± 0.27	0.49 ± 0.22