Table 4.
Performance of transformer model for EEG signals with eyes open on different methods for three class (ADHD, MDD, and SMC) classification
| Decision | Method | Accuracy | F1-score | Precision | Recall |
|---|---|---|---|---|---|
| Window-level | Trans + CCE | 71.87 | 66.57 | 65.57 | 71.56 |
| Trans + Focal Loss ( = 2) | 71.43 | 62.10 | 68.09 | 59.67 | |
| Trans + Focal Loss ( = 0.5) | 75.11 | 63.87 | 82.62 | 60.25 | |
| Trans + WCCE | 67.86 | 64.01 | 61.73 | 72.42 | |
| Patient-level | Trans + CCE | 68.75 | 60.01 | 60.37 | 61.87 |
| Trans + Focal Loss ( = 2) | 73.44 | 63.96 | 74.84 | 60.36 | |
| Trans + Focal Loss ( = 0.5) | 76.56 | 65.75 | 91.07 | 60.71 | |
| Trans + WCCE | 70.31 | 67.04 | 64.78 | 74.39 |
The best performance is in bold