Table 2 Performance comparison of methods trained/tested on DRIVE, CHASE-DB, and HRF.

From: State-of-the-art retinal vessel segmentation with minimalistic models

 

# Pub/Year

# Params

DRIVE

CHASE-DB

HRF

AUC

Dice

MCC

AUC

Dice

MCC

AUC

DICE

MCC

Maninis et al.24

ECCV/2016

82.20

Zhang et al.6

TMI/2016

96.36

96.06

96.08

74.10

Fu et al.25

MICCAI/2016

94.04

78.75

94.82

75.49

Liskowski et al.23

TMI/2016

48,000,000

97.90

98.45

Orlando et al.22

TBME/2017

95.07

78.57

75.56

95.24

73.32

70.46

95.24

71.58

68.97

Gu et al.52

TMI/2017

78.86

75.89

72.02

69.08

77.49

75.41

Wu et al.53

MICCAI/2018

98.07

98.25

Yan et al.31

TBME/2018

97.52

81.83

97.81

78.14

Wang et al.54

BSPC/2019

81.44

78.95

78.63

76.55

Wang et al.55

MICCAI/2019

97.72

82.70

98.12

80.37

Araujo et al.56

MICCAI/2019

97.90

98.20

Fu et al.57

MICCAI/2019

97.19

80.48

Wang et al.58

PatRec/2019

80.93

78.51

78.09

75.91

77.31

Wu et al.59

TMI/2019

97.79

Zhao et al.39

TMI/2019

78.82

76.59

Laibacher et al.44

CVPR-W/2019

549,748

97.14

80.91

97.03

80.06

78.14

Shin et al.26

MedIA/2019

7,910,000

98.01

82.63

98.30

80.34

98.38

81.51

Zhao et al.35

PatRec/2020

82.29

77.31

Zhuo et al.50

CMPB/2020

97.54

81.63

Mou et al.34

TMI/2020

56,030,000

97.96

98.12

Little U-Net

 

34,201

97.98

82.41

79.81

98.22

80.29

78.23

98.11

80.59

78.60

Little W-Net

 

68,482

98.10

82.79

80.24

98.47

81.69

79.74

98.25

81.03

79.09

  1. Best results are marked bold. A result is underlined whenever it lies within the confidence interval of the Little W-Net model (specified in Table 3 below).