Table 2 Selected radiomic machine learning model and clinical-radiological-radiomic machine learning model performance in training and validation set
From: MRI-based machine learning models predict the malignant biological behavior of meningioma
Models (Sequence-ROI-FS-ML) |  | Training set (n = 224) | Validation set (n = 89) |  | ||||
|---|---|---|---|---|---|---|---|---|
| Â | CV-AUC (95%CI) | CV-RSD | AUC (95%CI) | Sensitivity (95%CI) | Specificity (95%CI) | Delong test | ||
Z-score | p | |||||||
Grade (T1CE-2D-LASSO-LR) | Radiomic | 0.857 (0.836–0.880) | 0.060 | 0.829 (0.786–0.863) | 0.815 (0.613–0.930) | 0.661 (0.529–0.774) | -1.424 | 0.154 |
CRR |  |  | 0.821 (0.759–0.858) | 0.778 (0.573–0.906) | 0.629 (0.497–0.746) |  |  | |
Ki-67 (T1CE-3D-LASSO-NB) | Radiomic | 0.798 (0.745–0.854) | 0.090 | 0.752 (0.693–0.776) | 0.700 (0.504–0.846) | 0.780 (0.649–0.873) | 0.073 | 0.942 |
CRR |  |  | 0.753 (0.692–0.782) | 0.733 (0.538–0.870) | 0.763 (0.631–0.860) |  |  | |
Grade & Ki-67 (T1CE-2D-LASSO-LR) | Radiomic | 0.888 (0.856–0.923) | 0.051 | 0.904 (0.876–0.914) | 0.927 (0.790–0.981) | 0.604 (0.453–0.739) | 0.233 | 0.816 |
CRR |  |  | 0.906 (0.876–0.916) | 0.878 (0.730–0.954) | 0.708 (0.557–0.826) |  |  | |