TABLE IV.
THE AUCS OF CLASSIFIERS ON TSS CLASSIFICATION BOLD INDICATED THE HIGHEST AUC FOR A GIVEN CLASSIFIER
| Classifier | Ho, et al. (2017) [26] | No AE | AIF coupling patch only | Contralateral coupling patch only | AIF + contralateral patch | |||
|---|---|---|---|---|---|---|---|---|
| BI† | AE¥ | BI+AEΩ | AE | BI+AE | AE | BI+AE | ||
| LR | 0.574 | 0.618 | 0.650 | 0.658 | 0.647 | 0.676 | 0.710 | 0.765* |
| RE | 0.624 | 0.640 | 0.650 | 0.669 | 0.662 | 0.682 | 0.592 | 0.690 |
| GBRT | 0.567 | 0.608 | 0.590 | 0.570 | 0.676 | 0.674 | 0.612 | 0.670 |
| SVM | 0.669 | 0.636 | 0.477 | 0.736 | 0.605 | 0.666 | 0.600 | 0.746* |
| SMR | 0.683 | 0.661 | 0.574 | 0.707 | 0.650 | 0.677 | 0.705 | 0.730 |
BI = Models were trained with the baseline imaging features (94 descriptive statistics and 8 morphological features)
AE = Models were trained with the deep AE features (384 descriptive statistics generated from deep AE feature maps)
BI + AE = Models were trained with the baseline and deep AE features