Table 6.
Comparison of performance of the proposed approach against other algorithms for the atrial fibrillation (AFIB) classification problem on the 2017 PhysioNet/CinC Challenge dataset.
| Work | Approach | Per-class Performance (F1%) |
Overall Performance |
||||
|---|---|---|---|---|---|---|---|
| N | A | O | ~ | MF1 | Accuracy | ||
| ELP | CNN | 82.26 | 63.47 | 56.69 | 55.18 | 64.40 | 72.62 |
| ELP | RNN | 79.88 | 56.06 | 44.32 | 43.31 | 55.89 | 67.66 |
| ELP | RNN-Attention | 83.98 | 64.57 | 55.84 | 52.58 | 64.24 | 74.22 |
| Andreotti et al. [2] | Deep residual CNN | 82.6 | 46.6 | 60.0 | 60.2 | 62.4 | - |
MF1: Macro-averaging of F1-score