0244 Validation of Spectral Scoring with Polysomnography in Pediatric Sleep

Abstract Introduction Sleep scoring algorithms have advanced efficiency in evaluating sleep staging, however they have primarily been developed for adult populations. The present preliminary study seeks to evaluate whether a spectral scoring approach, previously validated in adults, is also valid in pediatric populations. While this approach was designed to capture dominant spectral patterns, rather than replicate specific visual stages, visual-spectral comparisons are essential to evaluate how these methods correspond. Methods Participants (n = 22, 9-13 years, 15F) completed overnight sleep studies monitored by polysomnography at the Children’s Hospital of Orange County. Visual scoring was determined by registered polysomnographic technologists and confirmed by board certified sleep physicians following AASM guidelines. Spectral activity within each epoch was assessed for the following: Wake (40–95 Hz), Light (∼11– 15.5 Hz), Hi Deep (1–3 Hz), Lo Deep (0.1–1 Hz), and REM (∼17–26 Hz). For each 30-sec epoch, the conditional probability of the participant being in each stage was calculated and the stage with maximal probability was assigned as the spectral score. Paired t-tests compared time spent in each stage for visual vs. spectral scores. Visually scored stage 2 (N2) was compared to spectral Light sleep. Visually scored slow wave sleep (SWS) was compared to spectral Lo Deep. Visually scored REM and total sleep time (TST) were compared to their spectrally scored equivalents. Results Overall, time in sleep stages was consistent between visual and spectral scoring, with high similarity for time spent in REM sleep, SWS, and TST (e.g. no significant differences, p>.05). However, participants spent less time in spectrally scored Light Sleep compared to visually scored N2 (p=.04, t=-2.44). Conclusion These results provide preliminary evidence that spectral scoring is a promising method for efficiently scoring pediatric sleep. Ongoing analyses will further evaluate the discrepancy between visual Light Sleep and spectral N2. Given that this method assigns stages based on dominant spectral features within each epoch, this spectral approach provides additional, clinically relevant information that is not readily captured through traditional visual scoring. This may be particularly valuable for assessing sleep physiology in pediatric clinical populations, where developmental differences in EEG patterns may complicate visual scoring. Support (if any) CHOC Internal Grant and K08HD107161