Table 2.
Performance statistics of the automated artifact detector as compared to visual artifact scoring, when considering only the visually scored artifacts having total power greater than the 99% percentile.
| NREM epochs (N=86,250) | REM epochs (N=20,717) | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Th | N art | sens | spec | acc | PPV | NPV | K | N art | sens | spec | acc | PPV | NPV | K |
| 1 | 579 | 81.35 | 97.23 | 97.12 | 16.57 | 99.87 | 0.27 | 170 | 75.29 | 97.19 | 97.01 | 18.16 | 99.79 | 0.28 |
| 2 | 560 | 82.14 | 97.22 | 97.12 | 16.18 | 99.88 | 0.26 | 159 | 78.62 | 97.18 | 97.04 | 17.73 | 99.83 | 0.28 |
| 3 | 443 | 84.88 | 97.12 | 97.06 | 13.23 | 99.92 | 0.22 | 122 | 82.79 | 97.07 | 96.98 | 14.33 | 99.90 | 0.24 |
| 4 | 306 | 86.60 | 97.00 | 96.96 | 9.32 | 99.95 | 0.16 | 90 | 85.56 | 96.96 | 96.91 | 10.92 | 99.94 | 0.19 |
| 5 | 163 | 87.73 | 96.86 | 96.85 | 5.03 | 99.98 | 0.09 | 51 | 88.24 | 96.81 | 96.79 | 6.38 | 99.97 | 0.11 |
| 6 | 66 | 86.36 | 96.77 | 96.76 | 2.00 | 99.99 | 0.04 | 28 | 85.71 | 96.71 | 96.69 | 3.40 | 99.98 | 0.06 |
| 7 | 41 | 80.49 | 96.74 | 96.73 | 1.16 | 99.99 | 0.02 | 11 | 90.91 | 96.64 | 96.64 | 1.42 | 100.00 | 0.03 |
| 8 | 28 | 71.43 | 96.73 | 96.72 | 0.70 | 99.99 | 0.01 | 8 | 87.50 | 96.63 | 96.63 | 0.99 | 100.00 | 0.02 |
| 9 | 20 | 70.00 | 96.72 | 96.71 | 0.49 | 99.99 | 0.01 | 6 | 83.33 | 96.62 | 96.62 | 0.71 | 100.00 | 0.01 |
| 10 | 18 | 72.22 | 96.72 | 96.71 | 0.46 | 99.99 | 0.01 | 5 | 80.00 | 96.62 | 96.61 | 0.57 | 100.00 | 0.01 |
Th: sub-epoch threshold indicating the minimum number of visually identified artifactual 4-s sub-epochs that needed to be present within a 30-s epoch in order to consider this epoch as artifactual. N art: number of 30-s epochs that are considered artifactual with the corresponding sub-epoch threshold. Sens: sensitivity; spec: specificity; acc: accuracy; PPV: positive predictive value; NPV: negative predictive value; K: Cohen’s kappa. The automated pipeline detected 2843 epochs with artifact in NREM and 705 epochs with artifact in REM.