Skip to main content

Table 3 The results of subgroup analyses

From: Deep learning-based evaluation of panoramic radiographs for osteoporosis screening: a systematic review and meta-analysis

Subgroup

DL models

N (%)

Sen

(95%CI)

Spe

(95%CI)

PLR

(95%CI)

NLR

(95%CI)

DOR (95%CI)

Overall

21 (100)

0.80

(0.74–0.86)

0.92

(0.88–0.95)

10.67

(6.4–17.6)

0.21

(0.15–0.29)

50.42

(23–109)

DL Methods

      

AlexNet

3(14)

0.89

(0.73–0.96)

0.99

(0.49-1)

94

(-353-541)

0.11

(0.04–0.22)

822

(-3852-5497)

VGG

2(9.5)

0.90

(0.85–0.92)

0.81

(0.77–0.84)

4.73

(3.81–5.65)

0.13

(0.09–0.18)

35

(20–50)

ResNet

4(19)

0.67

(0.62–0.72)

0.92

(0.88–0.94)

8.27

(4.81–11.72)

0.36

(0.29–0.42)

23

(9–37)

EfficientNet

6(29)

0.86

(0.70–0.94)

0.94

(0.89–0.96)

13.57

(4.52–22.6)

0.15

(0.02–0.28)

89

(-47-227)

GoogleNet

2(9.5)

0.8

(0.78–0.83)

0.94

(0.66–0.99)

14

(-14-42.21)

0.2

(0.17–0.25)

67

(-76-210)

Other

4(19)

0.73

(0.68–0.79)

0.90

(0.71–0.97)

7.05

(-1.34-15.46)

0.3

(0.21–0.39)

22

(-11-58)

  1. Sen: sensitivity, Spe: specificity, PLR: positive likelihood ratios, NLR: negative likelihood ratios, DOR: diagnostic odds ratio, DL Methods: deep learning methods