Skip to main content

Table 2 Meta-analysis results of the C-index of the prediction model for KOA patients

From: Accuracy of machine learning in identifying candidates for total knee arthroplasty (TKA) surgery: a systematic review and meta-analysis

Modeling variables

Training set

Validation set

No

c-index

I2(%)

No

c-index

I2(%)

Clinical feature

4

0.713 (0.628 – 0.799)

99.7

2

0.656 (0.526 – 0.786)

96.4

Radiomics

 X-ray

3

0.895 (0.865 – 0.924)

84.9

12

0.882 (0.825 – 0.939)

98.2

 MRI

2

0.755 (0.508 – 1.000)

99.0

2

0.725 (0.499 – 0.950)

94.8

 Overall

5

0.841 (0.777 – 0.904)

97.4

14

0.861 (0.806 – 0.916)

98.1

Clinical feature + radiomics

 Clinical feature + X-ray

7

0.867 (0.824–0.909)

95.4

9

0.837 (0.810–0.865)

59.4

 Clinical feature + MRI

2

0.772 (0.539 – 1.000)

98.9

2

0.758 (0.540 – 0.977)

95.2

 Clinical feature + MRI + X-ray

6

0.842 (0.816–0.867)

86.7

2

0.876 (0.842–0.910)

0

 Clinical feature + US + X-ray

1

0.820 (0.725–0.915)

NA

   

 Overall

16

0.844 (0.815 – 0.873)

95.4

13

0.831 (0.799 – 0.863)

79.1

  1. KOA knee osteoarthritis, No. number of models, I2 heterogeneity, MRI magnetic resonance imaging, US ultrasound, NA not available