Skip to main content

Table 3 Segmentation results of different methods in the cross-center test scenario

From: HCA-DAN: hierarchical class-aware domain adaptive network for gastric tumor segmentation in 3D CT images

Training

Test

Method

DSC

JI

ASD

95HD

{D1, D2, D3}

D4

V-Net

51.02

36.64

12.03

35.21

3D FPN

52.46

38.05

13.27

30.60

nnU-Net

55.49

40.28

9.42

29.58

UNETR

54.68

39.52

12.35

33.71

Swin-Unet

56.42

41.76

9.26

27.28

CoTr

56.28

41.54

8.71

26.17

AsTr (Ours)

60.62

46.88

8.13

25.80

{D1, D2, D4}

D3

V-Net

35.47

24.34

15.47

40.04

3D FPN

39.86

28.75

14.62

40.19

nnU-Net

43.62

31.18

13.71

39.41

UNETR

44.12

31.71

13.92

38.75

Swin-Unet

45.33

33.06

11.78

35.62

CoTr

43.65

31.19

13.33

39.57

AsTr (Ours)

46.54

33.61

10.97

30.92

{D1, D3, D4}

D2

V-Net

43.65

30.82

18.95

38.67

3D FPN

42.75

29.78

20.32

39.41

nnU-Net

48.32

34.79

15.31

33.01

UNETR

46.85

33.12

18.34

36.98

Swin-Unet

47.25

33.60

16.54

36.52

CoTr

48.89

35.35

14.89

31.60

AsTr (Ours)

55.94

42.15

10.50

24.52

{D2, D3, D4}

D1

V-Net

52.35

38.96

12.81

28.69

3D FPN

53.67

40.22

10.60

25.96

nnU-Net

55.50

42.30

9.52

24.20

UNETR

53.58

40.15

11.62

27.38

Swin-Unet

55.69

42.51

10.98

24.5

CoTr

54.98

41.16

7.41

20.76

AsTr (Ours)

56.42

43.14

8.68

23.29