Table 6.
Performance comparision of transformer model with state-of-the-art for Raw EEG classification for five classes (ADHD, MDD, OCD, SMC, and Healthy)
| Author | Method | Accuracy | F1-score | Precision | Recall | # TP |
|---|---|---|---|---|---|---|
| Eye Open EEG | ||||||
| Lawhern et al. [31] | EEGNet | 54.89 | 36.53 | 42.04 | 41.13 | 53.86k |
| Schirrmeister et al. [32] | DeepConvNet | 59.49 | 35.77 | 37.26 | 37.99 | 207.53k |
| Proposed Method | Transformer | 63.21 | 41.99 | 42.51 | 41.49 | 72.64k |
| Eye Close EEG | ||||||
| Lawhern et al. [31] | EEGNet | 55.28 | 40.84 | 44.64 | 42.92 | 53.86k |
| Schirrmeister et al. [32] | DeepConvNet | 59.78 | 38.04 | 55.52 | 42.15 | 207.53k |
| Proposed Method | Transformer | 61.74 | 46.57 | 51.24 | 47.83 | 72.64k |
The best performance is in bold