Table 6:
A comparison with the latest sleep staging algorithms
| Work | Signals used | Database | Subjects (Recordings) | classes | k-fold | Ace | κ |
|---|---|---|---|---|---|---|---|
| Long et al. (2014) | RIP | part of SIESTA | 48 out of 584 | 4 classes | 10-fold | 64.9 | 0.41 |
| Fonseca et al. (2015) | ECG+RIP | part of SIESTA | 48 out of 584 | 4 classes | 10-fold | 69 | 0.49 |
| Tataraidze et al. (2016) | ECG+RIP | part of SHHSv1 | 625 out of 5793 | 4 classes | 5-fold | 71.4 | 0.57 |
| our work | ECG | SLPDB | 16 (18) | 4 classes | 10-fold | 75.4 | 0.54 |
| out work | ECG | SHHSv1 | 5793 | 4 classes | 10-fold | 65.9 | 0.47 |
| Fonseca et al. (2015) | ECG+RIP | part of SIESTA | 48 out of 584 | 3 classes | 10-fold | 80 | 0.56 |
| Long et al. (2014) | RIP | part of SIESTA | 48 out of 584 | 3 classes | 10-fold | 77.1 | 0.48 |
| Wei et al. (2018) | ECG | SLPDB | 16 (18) | 3 classes | 10-fold* | 77 | 0.56 |
| our work | ECG | SLPDB | 16 (18) | 3 classes | 10-fold | 81.6 | 0.63 |
| out work | ECG | SHHSv1 | 5793 | 3 classes | 10-fold | 75.3 | 0.57 |
| Thomas et al. (2005) | ECG | private | 35train+35test+15test | 2 classes | - | - | 0.439 |
| Fonseca et al. (2017) | ECG+RIP | part of SIESTA+private | 180 (342) | 2 classes | 10-fold | 78.71 | 0.55 |
| Yoon et al. (2017) | ECG | private | 26train+25test | 2 classes | - | 87.03 | 0.61 |
| our work | ECG | SLPDB | 16 (18) | 2 classes | 10-fold | 85.1 | 0.68 |
| out work | ECG | SHHSv1 | 5793 | 2 classes | 10-fold | 80.8 | 0.61 |
*
10-fold by epochs, not by recordings