Table 5 Comparison results of different segmentation algorithms in DRIVE,[Key: Best, Second Best].

From: TCDDU-Net: combining transformer and convolutional dual-path decoding U-Net for retinal vessel segmentation

Methods

F1

Acc

AUC

Sp

Se

Yu44

–

95.24

97.23

98.03

76.43

DUNet45

82.37

95.66

98.02

98.00

79.63

Yan46

–

95.38

97.50

98.20

76.31

U-Net++47

81.92

96.88

98.12

–

–

CTF-Net48

82.41

95.67

97.88

–

–

CcNet49

–

95.28

96.78

98.09

76.25

Yang30

–

95.79

–

97.51

83.53

Xu28

–

96.64

98.28

98.02

82.43

Ours (w/o background decoder)

82.48 ± 1.38

96.88 ± 0.23

98.66 ± 0.37

98.12 ± 0.47

84.21 ± 4.88

Ours (TCDDU-Net)

82.65 ± 1.57

96.98 ± 0.25

98.68 ± 0.37

98.38 ± 0.42

82.58 ± 5.18

  1. Significant values are in bold and underline.