Table 4 Comparison results of DSC (%) and HD95 (mm) metrics on the BraTS2021 Training set.

From: Generative adversarial DacFormer network for MRI brain tumor segmentation

Methods

DSC \(\uparrow\)

HD95 \(\downarrow\)

ET

WT

TC

MEAN

ET

WT

TC

MEAN

dResU-Net31

84.0

86.0

82.2

84.1

−

−

−

−

nnU-Net34

84.8

92.8

88.7

88.8

13.07

6.78

8.79

9.55

MMEF-nnUNet35

87.3

92.8

90.1

90.1

10.09

5.10

9.68

8.29

SPA-Net27

83.5

90.7

88.9

87.7

22.07

6.10

9.21

12.46

Residual U-Net32

78.2

90.3

83.9

84.2

26.14

8.69

11.57

15.46

TransBTS17

85.3

92.1

88.2

88.6

10.78

6.97

7.56

8.44

U-Netr14

86.2

92.5

91.8

90.2

11.28

7.74

7.85

8.96

nnFormer36

84.9

93.0

93.5

90.5

9.28

4.85

5.32

6.48

SSCFormer37

85.3

93.0

92.8

90.4

11.11

5.38

7.37

7.95

MedNeXt15

89.4

93.3

91.5

91.4

8.32

7.83

4.79

6.98

SRAU-Net33

89.6

91.1

88.3

89.7

−

−

−

−

GVAT-Net30

81.4

90.8

85.4

85.9

21.83

5.37

8.56

11.92

GDacFormer (Ours)

86.3

93.7*

93.5*

91.2*

8.41

5.54

4.93

6.29*

  1. *Denotes significant difference compared with SOTA by Wilcoxon signed rank test (p value < 0.05).