Table 4.
Mean (SD) of performance metrics for two best performing models using all sequences, individual sequences, and combined T1-CE and FLAIR sequences.
| Performance metric | Model | |||||
|---|---|---|---|---|---|---|
| LASSO full All Seqs |
Elastic net full All Seqs |
LASSO full F |
Elastic net full F |
adaBoost full CE + F |
LASSO full CE + F |
|
| Brier | 0.088 (0.036) | 0.088 (0.036) | 0.083 (0.042) | 0.088 (0.043) | 0.086 (0.040) | 0.102 (0.026) |
| Accuracy | 0.892 (0.061) | 0.892 (0.063) | 0.897 (0.054) | 0.885 (0.054) | 0.888 (0.063) | 0.887 (0.070) |
| ROC AUC | 0.953 (0.041) | 0.952 (0.038) | 0.951 (0.049) | 0.948 (0.049) | 0.951 (0.040) | 0.950 (0.042) |
| Sensitivity | 0.887 (0.086) | 0.893 (0.092) | 0.917 (0.064) | 0.903 (0.071) | 0.900 (0.080) | 0.907 (0.088) |
| Specificity | 0.897 (0.073) | 0.890 (0.079) | 0.877 (0.094) | 0.867 (0.102) | 0.877 (0.111) | 0.867 (0.115) |
LASSO (least absolute shrinkage and selection operator; Enet: elastic net; ada: boosting of classification trees with adaBoost; F: FLAIR; CE: T1-CE.