Table 7.
The results for the multi-branch DL fusion model with the clinical SEEG dataset using the cross-subject scheme (patients 1–5).
| Patients Set | Method | Extracted Feature | ACC | SE | SP |
|---|---|---|---|---|---|
| Pt1-4/Pt5 | Bi-LSTM-AM | 70 various features × 4segments | 83.28% | 84.40% | 82.18% |
| 1D-CNN | Automatic | 77.68% | 82.52% | 72.65% | |
| Proposed | 70 various features × 4 segments + Automatic |
87.59% | 88.58% | 86.62% | |
| Pt2-5/Pt1 | Bi-LSTM-AM | 70 various features × 4 segments | 85.16% | 86.32% | 84.03% |
| 1D-CNN | Automatic | 78.42% | 82.73% | 74.21% | |
| Proposed | 70 various features × 4 segments + Automatic |
87.68% | 87.16% | 88.21% | |
| Pt3-5,1/Pt2 | Bi-LSTM-AM | 70 various features × 4 segments | 88.59% | 82.96% | 91.92% |
| 1D-CNN | Automatic | 85.11% | 86.30% | 83.89% | |
| Proposed | 70 various features × 4 segments + Automatic |
90.30% | 89.06% | 91.58% | |
| Pt4-5,1-2/Pt3 | Bi-LSTM-AM | 70 various features × 4 segments | 86.20% | 83.08% | 89.41% |
| 1D-CNN | Automatic | 78.14% | 86.87% | 69.14% | |
| Proposed | 70 various features × 4 segments + Automatic |
88.93% | 88.42% | 89.41% | |
| Pt5,1-3/Pt4 | Bi-LSTM-AM | 70 various features × 4 segments | 83.82% | 86.09% | 81.63% |
| CNN | Automatic | 80.59% | 83.97% | 77.16% | |
| Proposed | 70 various features × 4 segments + Automatic | 85.63% | 87.91% | 83.42% |