Skip to main content

Table 2 Comparisons of registration results (mean ± std)

From: macJNet: weakly-supervised multimodal image deformable registration using joint learning framework and multi-sampling cascaded MIND

Methods

Tumor

Liver

Image

Time (s)

TRE (mm)

DSC (%)

Hd95 (mm)

TRE (mm)

DSC (%)

Hd95(mm)

MI (%)

SSIM (%)

Affined

7.38 ± 3.56

46.81 ± 32.67

7.70 ± 3.27

7.01 ± 2.95

90.94 ± 1.38

6.67 ± 1.51

32.11 ± 6.96

34.29 ± 10.54

22.5 ± 3.8

Elastix

5.23 ± 1.49

53.27 ± 23.47

7.13 ± 2.03

5.45 ± 2.80

93.54 ± 1.12

5.33 ± 1.22

34.67 ± 12.85

38.42 ± 10.32

84.1 ± 7.2

VoxelMorph

5.89 ± 3.17

50.95 ± 28.90

7.06 ± 1.93

6.83 ± 2.85

93.18 ± 1.28

5.79 ± 1.64

35.88 ± 6.74

36.83 ± 11.77

0.20 ± 0.01

LapIRN

5.48 ± 2.24

52.51 ± 22.64

7.03 ± 1.71

5.51 ± 1.39

93.64 ± 1.13

5.52 ± 1.37

42.03 ± 4.71

49.03 ± 12.66

0.17 ± 0.02

macJNet

5.05 ± 1.77

55.20 ± 18.77

6.71 ± 1.97

4.83 ± 1.49

94.75 ± 0.82

4.53 ± 1.11

44.12 ± 4.63

54.43 ± 11.62

0.18 ± 0.02

  1. Bold values indicate better results than other methods