Table 3.
Top ten best performing models using combination of T1-CE and FLAIR sequences ordered by ROC AUC.
| Combination of CE and F sequences | ||
|---|---|---|
| Model | Feature reduction | ROC AUC mean (SD) |
| ada | Full | 0.951 (0.040) |
| Lasso | Full | 0.950 (0.042) |
| GBRM | Corr | 0.950 (0.040) |
| Enet | Full | 0.948 (0.041) |
| MLP | Full | 0.948 (0.042) |
| GBRM | Full | 0.944 (0.043) |
| RF | Full | 0.943 (0.044) |
| ada | Corr | 0.941 (0.037) |
| RF | Corr | 0.939 (0.039) |
| SVMRad | Full | 0.936 (0.042) |
ada: boosting of classification trees with adaBoost; Lasso (least absolute shrinkage and selection operator; GBRM: generalized boosted regression model; Enet: elastic net; MLP: multi-layer perceptron; rf: random forest;; SVMRad: SVM with a radial kernel; full: full feature set; corr: High correlation filter; F: FLAIR; CE: T1-CE.