Table 2 Area under ROC for each model.

From: MIXTURE of human expertise and deep learning—developing an explainable model for predicting pathological diagnosis and survival in patients with interstitial lung disease

 

AUC

95% CI

Validation set

 Proposed model

 2.5×

0.68

0.54–0.83

 5×

0.9

0.81–0.99

 20×

0.9

0.81–0.99

 2.5× + 5×

0.88

0.78–0.98

 5× + 20×

0.92

0.85–1.00

 2.5× + 20×

0.89

0.80–0.98

 2.5× + 5× + 20×

0.92

0.84–1.00

 Non-Integrated model

 k = 4

0.52

0.37–0.68

 k = 8

0.65

0.50–0.81

 k = 10

0.49

0.33–0.65

 k = 20

0.47

0.31–0.63

 k = 30

0.61

0.46–0.76

 k = 50

0.56

0.40–0.72

 k = 80

0.52

0.36–0.68

Test set

 Proposed model

 2.5×

0.74

0.60–0.88

 5×

0.86

0.75–0.97

 20×

0.77

0.64–0.90

 2.5× + 5×

0.88

0.78–0.98

 5× + 20×

0.87

0.77–0.97

 2.5× + 20×

0.83

0.71–0.94

 2.5× + 5× + 20×

0.88

0.78–0.98

  1. AUC area under the receiver operator characteristic curve, CI confidence interval, k number of clusters.
Back to article page