Table 5 Comparison of the subject-specific ablation results using baseline algorithms on CHB-MIT dataset.
From: Synchronization-based graph spatio-temporal attention network for seizure prediction
Feature selection | Model | ACC | SEN | SPE | FPR | F1-score |
|---|---|---|---|---|---|---|
Spatial | GCN | 78.13 | 76.73 | 86.99 | 0.130 | 80.02 |
Spatial | GAT | 95.14 | 95.08 | 95.45 | 0.045 | 94.60 |
Temporal | Transformer | 94.34 | 97.86 | 97.28 | 0.027 | 96.85 |
Spatio-temporal | GCN + Transformer | 85.34 | 83.12 | 91.40 | 0.086 | 85.13 |
Spatio-temporal | GAT + LSTM | 92.24 | 92.82 | 94.06 | 0.059 | 91.66 |
Spatio-temporal | SGSTAN | 98.20 | 97.85 | 98.07 | 0.019 | 97.60 |
