Table 2.
Summary of the reviewed A phase classification procedures, showing only the article’s reported results from best-performing models. The table sorted the articles by classification approach (first thresholds, then conventional machine learning, followed by deep learning). When using the same approach, the results were sorted by performance. In the end, the mean and standard deviation of the metrics reported by at least two studies are presented. Abbreviations: bruxism (B), derivation (D), female (F), healthy (H), insomnia (I), male (M), narcolepsy (N), periodic leg movement (PLM), subjects (S), threshold-based (TB), with ages between (WAB), years old (YO)
| Article | Population | Classifier | Performance metrics (%) | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Subjects | Characteristics | Acc | Sen | Spe | AUC | Other | ||||||||||||||
| [32] | 41~ | S (8 H, 9 with I, 1 with B, 4 with SDB, and 19 with RBD) using F2-F4/Fp2-F4, F4-C4, C4-P4, P4-O2, C4-A1, and F4-A1 D | TB | - | 91.2 | - | - | Precision: 45.9; F1 score: 59.6, false negative rate: 8.8, false discovery rate: 54.1 | ||||||||||||
| [34] | 10* | H S (all M) using F4-C4 D | TB | - | 84.0 | 90.0 | - | Correctness: 77.0 | ||||||||||||
| [36] | 8^ | - | TB | - | 84.9 | 77.6 | - | Concordance: 81.1 | ||||||||||||
| [37] | 8 | H S (4 M and 4 F) using C3-A2 or the C4-A1 D | TB | 72.4 | 51.6 | 76.4 | - | - | ||||||||||||
| [39] | 6~ | H S (4 M and 2 F) using C3-A2 or the C4-A1 D | TB | 80.5 | 75.8 | 81.3 | - | - | ||||||||||||
| [35] | 10 | H S (5 M and 5 F) using F4-C4 D | TB | 83.5 | - | - | - | - | ||||||||||||
| [38] | 16~ | H S using C3-A2 or the C4-A1 D | TB | 86.1 | 67.0 | 89.6 | - | - | ||||||||||||
| [46] | 6~ | H S using C3-A2 or the C4-A1 D | SVM | 72.4 | 76.8 | 69.2 | 79.0 | F1 score: 69.9 | ||||||||||||
| [48] | 6~ | H S using C3-A2 or the C4-A1 D | Ensemble of trees | 73.4 | 76.6 | 71.0 | - | - | ||||||||||||
| [44] | 14 ~ WAB 23 and 78 YO | Using C3-A2 or the C4-A1 D | Linear DA | 75.0 | 78.0 | 74.0 | 76.0 | - | ||||||||||||
| [51] | 30 ~ WAG 14 and 67 YO (mean 31.0) | NFLE S (16 M and 14 F) using C4-A1 D | SVM | 76.0 | 79.0 | 76.0 | - | Weighted accuracy: 78.0 | ||||||||||||
| [45] | 14 ~ WAB 23 and 78 YO | S (9 H, 4 with SDB, and 1 with B) using C3-A2 or the C4-A1 D | FFNN | 79.0 | 76.0 | 80.0 | 78.0 | - | ||||||||||||
| [41] | 4 | H S (4 M and 2 F) using C3-A2 or the C4-A1 D | FFNN | 81.6 | 75.7 | 83.1 | - | - | ||||||||||||
| [42] | 4 | H S (4 M and 2 F) using C4-A1 and F4-C4 D | SVM | 84.1 | 73.8 | 85.9 | - | Cohen’s kappa: 50.0 | ||||||||||||
| [43] | 8 | H S (4 M and 4 F) using C3-A2 or the C4-A1 D | Linear DA | 84.9 | 72.5 | 86.6 | - | Cohen’s kappa: 45.0 | ||||||||||||
| [47] | 77 ~ WAB 16 and 82 YO (mean 48.0) | S (6 H, 7 with I, 5 with N, 27 with NFLE, 9 with PLM, 22 with RBD, and 1 with SDB) using C4-A1 and F4-C4 D | Ensemble of trees | 78.0 | 74.0 | - | 86.0 | Precision: 80.0, Cohen’s kappa: 56.0 | ||||||||||||
| [59] | 6~ | H S using C3-A2 or the C4-A1 D | CNN | 53.0 | 92.1 | - | - | Precision: 20.1, F1 score: 33.0 | ||||||||||||
| [58] | 6~ | H S using C3-A2 or the C4-A1 D | CNN | 60.6 | 69.5 | 59.3 | 71.1 | Precision: 19.8, F1 score: 30.8 | ||||||||||||
| [53] | 14~ | D (9 H, 4 with SDB, and 1 with B) using C3-A2 or the C4-A1 D | DAE | 67.0 | 55.0 | 69.0 | - | - | ||||||||||||
| [56] | 15 ~ WAB 23 and 42 YO (mean 32.2) | H S (6 M and 9 F) using C3-A2 or the C4-A1 D | LSTM | 69.7 | 51.2 | 81.1 | 66.3 | - | ||||||||||||
| [54] | 19 ~ WAB 23 and 78 YO (mean 40.6) | S (15 H and 4 with SDB, 10 M and 9 F) using C3-A2 or the C4-A1 D | LSTM | 76.0 | 74.6 | 76.6 | 75.2 | Positive predictive value: 65.9, negative predictive value: 82.1, diagnostic odds ratio: 9.6 | ||||||||||||
| [60] | 9 ~ WAB 23 and 42 YO (mean 32.2) | H S (5 M and 4 F) using C3-A2 or the C4-A1 D | CNN | 76.3 | - | - | - | - | ||||||||||||
| [55] | 16 ~ WAB 16 and 67 YO (mean 32.9) | S (8 H and 8 with NFLE, 5 M and 11 F) using Fp23-F4, F4-C4, and the C4-A1 D | LSTM | 76.5 | 72.9 | 77.1 | 82.4 | - | ||||||||||||
| [57] | 16~ | H S using C3-A2 or the C4-A1 D | LSTM | 82.4 | 75.3 | 83.9 | - | F1 score: 57.4 | ||||||||||||
| [65] | 19~ | S (15 H and 4 with SDB, 10 M and 9 F) using C3-A2 or the C4-A1 D | LSTM | 83.0 | 76.5 | 83.4 | 88.2 | - | ||||||||||||
| [63] | 15~ | H S using C3-A2 or the C4-A1 D | LSTM | 83.0 | 76.1 | 84.2 | - | F1 score: 58.2 | ||||||||||||
| [62] | 15 ~ WAB 23 and 42 YO (mean 32.4) | H S (6 M and 9 F) using C3-A2 or the C4-A1 D | Ensemble of CNNs | 83.3 | 80.1 | 83.7 | - | Average: 82.4 | ||||||||||||
| [61] | 16~ | H S using C3-A2 or the C4-A1 D | CNN | 92.5 | 63.6 | 96.1 | 79.8 | Precision: 64.5, F1 score: 62.4 | ||||||||||||
| Mean ± standard deviation | 14.4 ± 11.3 S | TB | 80.6 ± 5.1 | 76.5 ± 12.4 | 82.7 ± 5.3 | - | - | |||||||||||||
| 18.1 ± 22.2 S | Conventional machine learning | 78.3 ± 4.3 | 75.8 ± 2.0 | 78.2 ± 6.2 | 79.8 ± 3.8 | - | ||||||||||||||
| 13.8 ± 4.3 S | Deep learning | 75.3 ± 10.6 | 71.5 ± 11.0 | 79.4 ± 9.4 | 77.2 ± 7.2 | - | ||||||||||||||