Skip to main content

Table 4 Diagnostic evaluation indices in the deep learning group and endoscopist group before propensity score matching in subgroups for the severity of CAG

From: Deep learning as a novel method for endoscopic diagnosis of chronic atrophic gastritis: a prospective nested case–control study

 

Mild CAG versus CNAG (104 vs. 793)

Moderate CAG versus CNAG (147 vs. 793)

Severe CAG versus CNAG (87 vs. 793)

DL

Endoscopist

DL

Endoscopist

DL

Endoscopist

Sensitivity

72.12%

39.42%

85.71%

62.59%

95.40%

90.80%

Specificity

96.34%

80.45%

96.34%

80.45%

96.34%

80.45%

PV+

72.12%

20.92%

81.29%

37.25%

74.11%

33.76%

PV−

96.34%

91.01%

97.32%

92.06%

99.48%

98.76%

Accuracy

93.53%

75.70%

94.68%

77.66%

96.25%

81.48%

Youden index

68.46%

19.87%

82.05%

43.04%

91.74%

71.25%

Odd product

68.13

2.68

158.07

6.89

546.66

40.65

LR+

19.70

2.02

23.42

3.2

26.07

4.64

LR−

0.29

0.75

0.15

0.47

0.05

0.11

  1. DL deep learning, PV+ positive predictive value, PV− negative predictive value, LR+ positive likelihood ratio, LR− negative likelihood ratio