Table 4 Recognition rates and kappa coefficient for different sizes of codebook by DHMM using ten testing subjects
From: A transition-constrained discrete hidden Markov model for automatic sleep staging
codebook = 30 | DHMM (M SSRR : 83.39% andK = 0.69) | |||||
|---|---|---|---|---|---|---|
Manual scoring | Â | Wake | S1 | S2 | SWS | REM |
| Â | Wake (%) | 81.11 | 8.87 | 7 | 0.09 | 2.93 |
| Â | S1 (%) | 24.38 | 26.6 | 31.08 | 3.96 | 13.97 |
| Â | S2 (%) | 4.77 | 1.24 | 77.03 | 11.58 | 5.39 |
| Â | SWS (%) | 0.55 | 0 | 3.54 | 95.92 | 0 |
| Â | REM (%) | 0 | 1.72 | 2.94 | 0.49 | 94.84 |
codebook = 40 | DHMM ( M SSRR : 84.36% K  = 0.71) | |||||
Manual scoring | Wake (%) | 86.82 | 4.29 | 6.63 | 0.09 | 2.17 |
| Â | S1 (%) | 29.17 | 30.31 | 28.59 | 0.20 | 11.72 |
| Â | S2 (%) | 2.94 | 1.14 | 82.41 | 10.95 | 4.35 |
| Â | SWS (%) | 0.31 | 0 | 4.55 | 95.14 | 0 |
| Â | REM (%) | 0 | 5.85 | 5.33 | 0.11 | 88.7 |
codebook = 50 | DHMM ( M SSRR : 85.29% K  = 0.73) | |||||
Manual scoring | Wake (%) | 88.81 | 2.64 | 6.29 | 0.09 | 2.17 |
| Â | S1 (%) | 21.44 | 33.62 | 33.19 | 0.21 | 11.54 |
| Â | S2 (%) | 2.16 | 1.08 | 81.58 | 11.46 | 3.72 |
| Â | SWS (%) | 0.23 | 0 | 4.85 | 94.92 | 0 |
| Â | REM (%) | 0 | 3.47 | 5.05 | 1.34 | 90.14 |
codebook = 60 | DHMM ( M SSRR : 85.12% K  = 0.69) | |||||
Manual scoring | Wake (%) | 83.68 | 7.57 | 5.82 | 0.08 | 2.84 |
| Â | S1 (%) | 26.41 | 26.71 | 33.2 | 0 | 13.67 |
| Â | S2 (%) | 2.66 | 1.66 | 82.3 | 9.82 | 3.56 |
| Â | SWS (%) | 0 | 0 | 7.24 | 92.76 | 0 |
| Â | REM (%) | 1.23 | 2.3 | 6.07 | 0 | 90.4 |