TABLE I:

Intra-dataset evaluation of the fully-supervised methods. Bold values indicate the best results. Our method significantly outperformed most of the competing methods, except for those underlined entries (p>0.05).

Methods	UCLA dataset				NIH dataset
	DSC [%]	ASD [mm]	ASD-shadow [mm]	HD [mm]	DSC [%]	ASD [mm]	ASD-shadow [mm]	HD [mm]
Radial-2.5D-UNet (2020) [20]	88.56(2.78)	1.46(0.41)	1.81(0.72)	7.21(2.16)	86.13(5.49)	1.80(1.07)	2.33(1.78)	8.38(4.08)
VNet (2016) [21]	91.78(2.43)	0.99(0.33)	1.16(0.53)	6.02(1.97)	88.15(4.57)	1.51(0.78)	2.26(1.87)	7.86(4.22)
UNet (2015) [8]	91.96(2.38)	0.98(0.34)	1.16(0.54)	6.22(2.33)	89.28(4.55)	1.34(0.68)	1.92(1.28)	7.36(3.65)
nnUNet (2021) [18]	92.17(2.21)	0.94(0.31)	1.08 (0.50)	5.79 (2.08)	89.36(4.56)	1.35(0.81)	1.94 (1.53)	7.09 (3.97)
DAF-Net (2019) [11]	91.96(2.35)	0.97(0.32)	1.12(0.53)	5.80 (1.95)	89.07(4.15)	1.39(0.74)	2.02(1.56)	6.92 (3.59)
SCO-SSL (semi-supervised)	91.60(2.37)	1.02(0.34)	1.22(0.52)	6.37(2.36)	90.12 (3.61)	1.23 (0.63)	1.80 (1.18)	6.65 (2.89)
SCO-SSL (full-supervised)	92.25 (2.19)	0.93 (0.29)	1.10(0.46)	5.89(1.93)	89.85(3.30)	1.26(0.58)	1.84(1.13)	6.88(3.00)