Table 3.
Performance of radiomics models using three machine learning classifiers.
| Model | Cohort | Accuracy | AUC (95% CI) | Sensitivity | Specificity | PPV | NPV | F1-score | Threshold |
|---|---|---|---|---|---|---|---|---|---|
| LR | Training | 0.866 | 0.913 (0.879–0.948) | 0.895 | 0.782 | 0.922 | 0.723 | 0.908 | 0.706 |
| LR | Validation | 0.878 | 0.921 (0.872–0.969) | 0.883 | 0.860 | 0.953 | 0.698 | 0.917 | 0.682 |
| ET | Training | 0.880 | 0.885 (0.844–0.926) | 0.914 | 0.782 | 0.923 | 0.761 | 0.919 | 0.663 |
| ET | Validation | 0.806 | 0.898 (0.839–0.957) | 0.766 | 0.930 | 0.972 | 0.556 | 0.857 | 0.754 |
| MLP | Training | 0.882 | 0.943 (0.917–0.970) | 0.886 | 0.873 | 0.952 | 0.727 | 0.918 | 0.787 |
| MLP | Validation | 0.917 | 0.912 (0.852–0.973) | 0.964 | 0.767 | 0.930 | 0.868 | 0.946 | 0.431 |
LR, Logistic Regression; ET, Extremely Randomized Trees;MLP, Multi-layer Perceptron; AUC, area under the curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value.