Bedtime Brain State Predicts the Impact of Closed-Loop Auditory Stimulation on Sleep and Cognition

Sleep interventions targeting slow-wave activity (SWA) show heterogeneous effects across individuals. We investigated whether pre-sleep brain states predict responsiveness to auditory stimulation and associated cognitive benefits. Twenty-eight healthy adults (19-40 years) completed three overnight laboratory visits under mild sleep restriction. Pre-sleep EEG recordings captured spectral band power before each night. Participants received auditory stimulation or sham on two randomized nights with polysomnography monitoring. Memory recall and sustained attention were assessed at multiple timepoints the following day. Machine learning models using normalized EEG features and transfer learning with pre-trained sleep architectures (SleepNet, DeepSleepNet, TinySleepNet) predicted responsiveness. Auditory stimulation significantly enhanced SWA (p < 0.01) in over 90% of participants. Pre-sleep alpha and theta power correlated with SWA enhancement (alpha: r = 0.436; theta: r = 0.329; both p < 0.01), as did sleep onset latency (r = 0.429, p < 0.01). Temporal embeddings derived from 8 minutes of pre-sleep EEG before lights off predicted individual responsiveness to auditory stimulation with 80% accuracy (AUC = 0.93). Pre-sleep brain state determines responsiveness to auditory stimulation, with individuals exhibiting longer sleep onset latency showing greater SWA enhancement. Tailoring pre-sleep interventions based on these predictive features may optimize brain receptivity to auditory stimulation during sleep.