Novel Real-Time Algorithms:Quantifying Wake-Sleep States

点击翻译按钮获取中文摘要

The overall goal of this project is to develop and clinically validate a novel class of computerized EEG-based algorithms and sophisticated signal processing methods designed for high temporal resolution characterization of the wake-sleep transition and fluctuations within sleep stages. In addition to their capability of profiling wakefulness and various sleep stages as a continuous process, these algorithms will be developed in a format such that they can be eventually implemented for real-time detection of wake-to sleep transitions. The new approach of this proposal is based on an innovative interaction of classically independent signal processing methods for quantifying EEG, including traditional Short-Term Fourier Transform, Recursive Auto-regressive Parameter Estimation, and Wavelet analysis. This new approach takes advantage of each method's relative strength and "fuses" the information obtained from each technique to provide a means of quantifying the entire wake-sleep transition with high temporal resolution (e.g., early detection of sleep onset) while being robust to factors that normally degrade EEG processing. The developed algorithms will be validated with an extensive set of clinical data including nocturnal polysomnography, and day-time tests of Sleepiness, MSLT (Multiple Sleep Latency Test) and MWT (Maintenance of Wakefulness Test). The dynamic events detected by these algorithms (e.g., micro-arousals and micro-sleeps) will be evaluated against standard R&K scoring of the same EEG data. Furthermore, the profile and characteristics of the events detected in PSG will be correlated with the daytime objective and subjective tests of sleepiness.