Table 4 The test results of the compare study in polypGen.

From: Dual ensemble system for polyp segmentation with submodels adaptive selection ensemble

Train/Val

(C2, C3, C4, C5, C6) 90%/10%

(C1, C2, C4, C5, C6) 90%/10%

(C1, C2, C4, C5, C6) 90%/10%

Test

C1

C2

C3

Metric

Dice

mIoU

HD95

p

Dice

mIoU

HD95

p

Dice

mIoU

HD95

p

UNet

0.737

0.805

35.62

< 0.05

0.714

0.785

36.42

< 0.05

0.742

0.812

34.18

< 0.05

UperNet

0.818

0.853

23.05

< 0.05

0.806

0.832

28.54

< 0.05

0.820

0.856

21.64

< 0.05

Double-UNet

0.672

0.749

43.23

< 0.05

0.658

0.726

53.12

< 0.05

0.682

0.762

39.54

< 0.05

SS-Former

0.837

0.865

18.39

< 0.05

0.820

0.874

19.87

0.26

0.894

0.910

9.08

0.09

FCB-Former

0.848

0.875

17.65

< 0.05

0.803

0.862

23.32

0.08

0.902

0.916

9.06

0.20

ESFPNet

0.846

0.873

17.92

< 0.05

0.836

0.881

18.09

0.69

0.898

0.912

9.33

0.11

HarDNet-DFUS

0.831

0.867

18.82

< 0.05

0.813

0.869

21.67

0.19

0.852

0.879

12.61

< 0.05

DET-Former

0.872

0.894

12.91

_

0.843

0.886

17.37

_

0.912

0.922

7.63

_

Back to article page