Automatic sleep stage detection using consecutive and non-consecutive approach for elderly and young healthy subject

The difference in sleep stages between young, elderly and different-gendered healthy or not, has become a significant concern in sleep medicine. Age has been known to have a significant affect on sleep structure. Current methods of sleep stage detection, such as automatic sleep stage detection, are becoming an important tool when examining the quality of sleep. In this study, we first evaluated a new approach for automatic sleep stage detection based on three consecutive and non-consecutive 1-second sub-epochs of EEG signal. We then compared the effectiveness of using a fixed-duration 5-second epoch and three consecutive and non-consecutive 1-second sub-epochs. Finally, this study investigated which method is most effective in detecting the sleep stages of ten healthy female subjects from various age groups. Several measurements were extracted from the EEG signal, including the Relative Band Spectral Energy (RBSE), Power Rations (PR), Central Frequency (CF), Spectral Edge Frequency (SEF), Root Mean Square Frequency (RMSF) and Spectral Entropy). Study results demonstrate that the proposed technique was able to detect sleep stages with epoch-by-epoch agreements of 91% for M3 (3 non-consecutive 1 sec sub-epoch method), 82% for M2 (3 consecutive 1sec sub-epoch method) and 70% for M1 (5 sec epoch).