Table 2.
Diagnostic performance of the clinical, radiomics, and combined models.
| Model | AUC (95%CI) | Sensitivity | Specificity | Accuracy | p-value |
|---|---|---|---|---|---|
| Training set | |||||
| Clinical model (LR) | 0.769 (0.716 to 0.817) | 0.50 | 0.91 | 0.71 | – |
| Clinical model (SVM) | 0.871 (0.827 to 0.908) | 0.66 | 0.93 | 0.80 | <0.001 |
| Radiomics model 1 | 0.951 (0.919 to 0.973) | 0.84 | 0.92 | 0.88 | <0.001 |
| Radiomics model 2 | 0.991 (0.972 to 0.999) | 0.96 | 0.97 | 0.96 | <0.001 |
| Combined model (LR) | 0.950 (0.918 to 0.972) | 0.85 | 0.94 | 0.90 | <0.001 |
| Combined model (SVM) | 0.978 (0.854 to 0.992) | 0.92 | 0.97 | 0.95 | <0.001 |
| Test set | |||||
| Clinical model (LR) | 0.781 (0.684 to 0.860) | 0.53 | 0.89 | 0.77 | – |
| Clinical model (SVM) | 0.744 (0.644 to 0.828) | 0.69 | 0.85 | 0.80 | 0.539 |
| Radiomics model 1 | 0.924 (0.850 to 0.968) | 0.78 | 0.89 | 0.85 | 0.012 |
| Radiomics model 2 | 0.936 (0.866 to 0.976) | 0.78 | 0.90 | 0.86 | 0.013 |
| Combined model (LR) | 0.933 (0.862 to 0.974) | 0.75 | 0.85 | 0.82 | 0.006 |
| Combined model (SVM) | 0.928 (0.856 to 0.971) | 0.81 | 0.87 | 0.85 | 0.009 |
The p-value was calculated by the Delong test. AUC, area under the receiver operating characteristic curve; CI, confidence interval; SVM, support vector machine; LR, logistic regression. Radiomics model 1, the best radiomics model based on a single phase (arterial-LASSO-SVM). Radiomics model 2, the best radiomics model based on the multiphasic phase (three-LASSO-SVM).