Research Article
DWT-Net: Seizure Detection System with Structured EEG Montage and Multiple Feature Extractor in Convolution Neural Network
Table 4
Classifier performance on TUSZ.
| Classifier | Sensitivity | Specificity | TPR |
| 5 sec window origin dataset | 45.69% | 76.06% | 23.11% | 1 sec window origin dataset | 35.86% | 83.25% | 28.59% | 5 sec window 5-fold | 77.23% | 83.41% | 37.78% | 1 sec window 5-fold | 79.10% | 81.38% | 34.61% | 5 sec window with +1 conv layers | 76.28% | 80.91% | 34.90% | 5 sec window with +2 conv layers | 77.82% | 81.33% | 34.35% | 5 sec window without SMOTE | 67.89% | 82.45% | 33.49% | 5 sec window double banana montage | 70.98% | 73.17% | 33.34% |
|
|