Pre‑sleep central-occipital alpha-theta dynamics predict NREM2 onset in adult.

Abstract

Sleep onset (SO) marks the transition from wakefulness to a restorative state of sleep in the human brain. Recent studies suggest that the sleep onset process (SOP) is a back-and-forth progression that occurs before the brain reaches Non-Rapid Eye Movement Sleep Stage 2 (NREM2) and that its duration is highly variable. Very short sleep onset latencies may increase the risk of accidents, while long SO may indicate poor sleep quality and insomnia. In this study, we investigated EEG power dynamics before the first NREM2 sleep epoch to identify potential predictors of SO ending and factors that differentiate fast from slow sleepers. We found that both groups exhibited an initial rise followed by a decline in alpha rhythms, resembling an inverted-U curve, in central and occipital regions. Compared to slow sleepers, fast sleepers showed a narrower, steeper alpha-power curve and an earlier rise in theta power. When examining the alpha-theta ratio, we found that, while slow sleepers followed an inverted‑U-shaped progression in central and occipital regions, fast sleepers' ratio decreased monotonically toward NREM2. Finally, we combined the alpha-theta ratio and the temporal rate of decline in the ratio to predict NREM2 onset using supervised learning. A fuzzy C-means (FCM) classifier achieved the highest accuracy (0.65). These findings indicate that neurodynamics during SOP are directly related to its duration. The alpha-theta ratio in central and occipital channels, together with its rate of decline, provides a compact and physiologically interpretable predictor of NREM2 onset. This predictive capability supports applications in both occupational safety and sleep health by enabling early detection of unintended sleep onset and allowing individuals to adopt proactive interventions when a prolonged SOP is anticipated.