Table 3.
Diagnostic performance of the radiomics models.
| Model | SVM | LR | |||
|---|---|---|---|---|---|
| Training | Test | Training | Test | ||
| Arterial phase | |||||
| LASSO | |||||
| AUC (95% CI) | 0.951 (0.919 to 0.973) | 0.924 (0.850 to 0.968) | 0.936 (0.901 to 0.961) | 0.919 (0.845 to 0.966) | |
| Sensitivity | 0.84 | 0.78 | 0.86 | 0.81 | |
| Specificity | 0.92 | 0.89 | 0.90 | 0.94 | |
| Accuracy | 0.88 | 0.85 | 0.88 | 0.89 | |
| ML | |||||
| AUC (95% CI) | 0.896 (0.855 to 0.929) | 0.901 (0.822 to 0.953) | 0.881 (0.838 to 0.916) | 0.896 (0.816 to 0.950) | |
| Sensitivity | 0.78 | 0.75 | 0.81 | 0.78 | |
| Specificity | 0.88 | 0.85 | 0.81 | 0.79 | |
| Accuracy | 0.83 | 0.82 | 0.81 | 0.79 | |
| RFE | |||||
| AUC (95% CI) | 0.909 (0.869 to 0.940) | 0.914 (0.838 to 0.962) | 0.904 (0.863 to 0.935) | 0.896 (0.816 to 0.950) | |
| Sensitivity | 0.81 | 0.84 | 0.80 | 0.72 | |
| Specificity | 0.90 | 0.87 | 0.87 | 0.84 | |
| Accuracy | 0.86 | 0.86 | 0.84 | 0.80 | |
| Venous phase | |||||
| LASSO | |||||
| AUC (95% CI) | 0.959 (0.929 to 0.979) | 0.920 (0.846 to 0.966) | 0.918 (0.880 to 0.947) | 0.906 (0.828 to 0.957) | |
| Sensitivity | 0.89 | 0.69 | 0.82 | 0.78 | |
| Specificity | 0.87 | 0.87 | 0.82 | 0.87 | |
| Accuracy | 0.88 | 0.81 | 0.82 | 0.84 | |
| ML | |||||
| AUC (95% CI) | 0.856 (0.810 to 0.894) | 0.831 (0.739 to 0.900) | 0.829 (0.780 to 0.871) | 0.807 (0.713 to 0.882) | |
| Sensitivity | 0.75 | 0.81 | 0.73 | 0.75 | |
| Specificity | 0.78 | 0.74 | 0.74 | 0.73 | |
| Accuracy | 0.77 | 0.77 | 0.73 | 0.73 | |
| RFE | |||||
| AUC (95% CI) | 0.926 (0.889 to 0.954) | 0.915 (0.839 to 0.963) | 0.909 (0.869 to 0.940) | 0.872 (0.787 to 0.932) | |
| Sensitivity | 0.80 | 0.72 | 0.81 | 0.69 | |
| Specificity | 0.89 | 0.95 | 0.83 | 0.87 | |
| Accuracy | 0.85 | 0.87 | 0.82 | 0.81 | |
| Plain scan | |||||
| LASSO | |||||
| AUC (95% CI) | 0.991 (0.972 to 0.998) | 0.909 (0.832 to 0.959) | 0.991 (0.972 to 0.998) | 0.892 (0.811 to 0.947) | |
| Sensitivity | 0.94 | 0.72 | 0.96 | 0.81 | |
| Specificity | 0.95 | 0.90 | 0.94 | 0.90 | |
| Accuracy | 0.94 | 0.84 | 0.95 | 0.87 | |
| ML | |||||
| AUC (95% CI) | 0.900 (0.859 to 0.932) | 0.899 (0.819 to 0.951) | 0.904 (0.864 to 0.935) | 0.848 (0.759 to 0.914) | |
| Sensitivity | 0.77 | 0.75 | 0.84 | 0.62 | |
| Specificity | 0.86 | 0.87 | 0.85 | 0.84 | |
| Accuracy | 0.81 | 0.83 | 0.84 | 0.77 | |
| RFE | |||||
| AUC (95% CI) | 0.933 (0.897 to 0.959) | 0.898 (0.819 to 0.951) | 0.921 (0.884 to 0.950) | 0.882 (0.799 to 0.939) | |
| Sensitivity | 0.81 | 0.78 | 0.85 | 0.78 | |
| Specificity | 0.90 | 0.92 | 0.84 | 0.87 | |
| Accuracy | 0.85 | 0.87 | 0.84 | 0.84 | |
| Three-phase combination | |||||
| LASSO | |||||
| AUC (95% CI) | 0.991 (0.972 to 0.999) | 0.936 (0.866 to 0.976) | 0.994 (0.977 to 1.000) | 0.931 (0.859 to 0.973) | |
| Sensitivity | 0.96 | 0.78 | 0.97 | 0.75 | |
| Specificity | 0.97 | 0.90 | 0.97 | 0.94 | |
| Accuracy | 0.96 | 0.86 | 0.97 | 0.87 | |
| ML | |||||
| AUC (95% CI) | 0.912 (0.873 to 0.942) | 0.903 (0.824 to 0.954) | 0.914 (0.875 to 0.943) | 0.890 (0.809 to 0.945) | |
| Sensitivity | 0.81 | 0.69 | 0.83 | 0.75 | |
| Specificity | 0.86 | 0.84 | 0.86 | 0.84 | |
| Accuracy | 0.84 | 0.79 | 0.85 | 0.81 | |
| RFE | |||||
| AUC (95% CI) | 0.908 (0.869 to 0.939) | 0.887 (0.805 to 0.943) | 0.909 (0.870 to 0.940) | 0.883 (0.800 to 0.940) | |
| Sensitivity | 0.81 | 0.69 | 0.82 | 0.72 | |
| Specificity | 0.89 | 0.85 | 0.85 | 0.89 | |
| Accuracy | 0.85 | 0.80 | 0.84 | 0.83 | |
AUC, area under the receiver operating characteristic curve; CI, confidence interval; SVM, support vector machine; LR, logistic regression; LASSO, least absolute shrinkage and selection operator; MI, mutual information; RFE, recursive feature elimination.