Table 1 Accuracy, sensitivity, and specificity for each method, organized by size of input each methods takes.

From: Skin strata delineation in reflectance confocal microscopy images using recurrent convolutional networks with attention

Input

Method

Accuracy

Sensitivity

Specificity

Epidermis

DEJ

Dermis

Epidermis

DEJ

Dermis

Whole stack

Toeplitz Att. (\(\text{D}=1\))

88.07

93.41

85.04

82.66

96.35

89.71

96.07

Toeplitz Att. (\(\text{D}=7\))

87.59

92.84

84.18

82.89

95.87

89.38

96.08

Toeplitz Att. (\(\text{D}=0\))

87.57

91.63

83.97

84.94

96.89

89.40

95.29

Toeplitz Att. (\(\text{D}=3\))

87.13

91.65

83.48

83.89

96.08

89.23

95.44

Toeplitz Att. (\(\text{D}=2\))

86.65

91.44

82.27

83.95

96.08

88.91

95.07

Toeplitz Att. (\(\text{D}=5\))

86.63

91.94

83.06

81.87

95.79

88.84

95.35

Global Att.

86.27

92.95

81.98

80.43

95.30

88.55

95.48

Full-Seq. (Bidir. GRU)29

87.97

93.95

83.22

84.16

95.82

90.54

95.51

Partial stack

Partial-Seq. (Bidir. GRU)29

87.52

94.14

82.54

83.33

94.78

90.83

95.44

IV3-Context29

86.95

92.00

83.52

82.64

95.61

89.24

95.56

Single image

Inception-V330

84.87

88.83

84.66

78.18

95.84

85.73

96.23

Hames et al.24

84.48

88.87

80.93

81.85

93.81

87.81

94.78

Kaur et al.25

72.12

82.44

62.75

66.09

89.48

79.35

89.18