Table 5 Metrics of the best performing ML models in feature selection.
| Â | AUC | Accuracy | Sensitivity | Specificity | Youden index |
|---|---|---|---|---|---|
(a) All patients | |||||
 Linear SVM | \(0.859\pm 0.005\) | \(0.800\pm 0.005\) | \(0.416\pm 0.013\) | \(\mathbf{0.958}\pm \mathbf{0.006}\) | \(0.374\pm 0.013\) |
 RBF SVM | \( \mathbf{0.865}\pm \mathbf{0.007}\) | \( \mathbf{0.856}\pm \mathbf{0.004}\) | \( \mathbf{0.621}\pm \mathbf{0.013}\) | \(0.953\pm 0.003\) | \( \mathbf{0.575}\pm \mathbf{0.013}\) |
 LASSO | \(0.859\pm 0.005 \) | \(0.802\pm 0.004\) | \(0.452\pm 0.014\) | \(0.946\pm 0.006\) | \(0.399\pm 0.012\) |
 Ridge | \(0.858\pm 0.005\) | \(0.796\pm 0.005\) | \(0.449\pm 0.014\) | \(0.939\pm 0.007\) | \(0.388\pm 0.012\) |
(b) Biopsy GG > 1 patients only | |||||
 Linear SVM | \( \mathbf{0.944}\pm \mathbf{0.004}\) | \(0.901\pm 0.005\) | \(0.637\pm 0.023\) | \(0.936\pm 0.004\) | \(0.573\pm 0.024\) |
 RBF SVM | \(0.894\pm 0.007\) | \(0.878\pm 0.004\) | \(0.445\pm 0.024\) | \(0.936\pm 0.004\) | \(0.381\pm 0.023\) |
 LASSO | \(0.930\pm 0.005\) | \(0.895\pm 0.005\) | \(0.613\pm 0.024\) | \(0.933\pm 0.005\) | \(0.545\pm 0.023\) |
 Ridge | \(\mathbf{0.944}\pm \mathbf{0.005}\) | \( \mathbf{0.904}\pm \mathbf{0.005}\) | \( \mathbf{0.652}\pm \mathbf{0.023}\) | \( \mathbf{0.938}\pm \mathbf{0.004}\) | \( \mathbf{0.590}\pm \mathbf{0.024}\) |
