0999 Characterizing Clinical Population Differences in Transient Oscillation Features in the Sleep EEG

• Published in: Sleep (New York, N.Y.) Vol. 41; no. suppl_1; p. A370
• Main Authors: Stokes, P A, Rath, P, Manoach, D S, Stickgold, R, Prerau, M J
• Format: Journal Article
• Language: English
• Published: US Oxford University Press 27.04.2018
• Subjects: Electroencephalography; Sleep
• ISSN: 0161-8105; 1550-9109
• DOI: 10.1093/sleep/zsy061.998

Abstract Introduction Sleep is traditionally characterized through patterns observed in the time-domain electroencephalogram (EEG). These include persistent oscillations, such as slow wave activity, or transient oscillatory activity, such as spindles and K-complexes. Time-domain identification of oscillations is difficult, with waveforms frequently being obscured by other oscillations and noise, limiting analyses to easily discernible waveforms. These difficulties lead to large inter-scorer variability—especially for transient oscillations like sleep spindles, the identification of which varies greatly depending on scorer or automated method used. This variability is a major challenge in analysis of sleep differences in clinical populations, such as patients with schizophrenia (SZ), in whom cognitive deficits and symptoms correlate with measurements of spindle density based on traditional techniques. Here we demonstrate a new, objective analysis method that is agnostic to arbitrary criteria and robust to time-domain obfuscations, and that may offer improved characterization of the dynamics of sleep, its natural variation, and biomarkers of disease. Methods We demonstrate a novel approach to the analysis of sleep EEG oscillations based on the observation that transient oscillations will appear as distinct peaks in the time-frequency spectrogram. By characterizing distributions of peak properties, rather than seeking to identify specific, pre-defined oscillations, we can perform definition-agnostic analyses and comparisons of transient oscillatory activity across subjects or groups. We applied this approach to full-night EEG recordings from 21 SZ patients and 17 healthy controls and compared the peak feature distributions between the groups. Results We found marked differences in the frequency distributions of time-frequency peaks between SZ and control participants, suggesting less frequent and less frequency-specific spindle-like peaks in SZ patients. The SZ distribution is also elevated over the control from 8-12Hz. The distribution of SZ peaks also suggests they are generally of smaller height than those of the control group. Conclusion These results demonstrate that transient oscillations can be robustly identified and objectively analyzed in this definition-agnostic approach. Furthermore, this approach could more comprehensively characterize differences in clinical populations and reduce method-based variability in spindle detection and other analyses. Support (If Any) R01NS096177; R01MH092638; K24MH099421.