Modeling Multi-scale Sleep EEG Dynamics as a Framework for Biomarker Discovery

 描述（由申请人提供）本研究的目的是开发多尺度睡眠EEG动力学的统计模型，其将形成用于识别心理和神经障碍的生物标志物的框架。睡眠是一种自然的、恢复性的、改变的意识状态，对每个活着的人都是如此。神经生理学上，睡眠是一个连续的动态过程，涉及大脑内皮层和皮层下网络在多个时间尺度上操作的复杂相互作用。因此，睡眠脑电图（EEG）的研究是一种理想的，自然的手段，同时观察相关的活动从许多大脑区域。毛皮，许多心理（例如，精神分裂症、抑郁症和焦虑症）和神经学（例如，阿尔茨海默氏病和帕金森氏病）与睡眠动力学紊乱有关，影响着全世界数百万人。尽管约2000万美国人患有睡眠动力学障碍，例如慢性和/或严重失眠，但目前的临床实践并不能客观地量化睡眠EEG动力学。这是因为目前的方法，虽然有助于我们目前的睡眠理解，限制了睡眠动态可以描述的程度-通过主观，视觉评分的做法离散的时间和状态的睡眠。此外，睡眠领域还没有采用强大的频谱估计技术，这可以大大提高睡眠EEG动力学的表征。因此，将最先进的动态建模和睡眠动态信号处理技术结合起来，开发客观的方法将有很大的好处。为了解决这个问题，我们最近开发了一个信号处理和建模框架来表征在不同时间尺度上演化的多个同时神经过程的动态。我们已经证明，这些动态模型在准确表征睡眠开始的行为和生理动力学方面明显优于传统的基于阶段的方法。此外，我们的初步研究表明，优化的谱估计方法极大地改善了健康受试者睡眠脑电动力学的分析，并揭示了阿尔茨海默病和精神分裂症患者病理性睡眠的生动特征。在这项提案中，我们将开发一种新的信号处理和动态建模框架，从一组来自健康和病理受试者的睡眠记录的大型数据库（约20，000条记录）中表征睡眠EEG动态。通过表征健康睡眠中的变异性，我们将开发一种新的异常检测方法，用于识别和量化病理人群睡眠EEG动力学的差异。

 DESCRIPTION (provided by applicant) The objective of this research is to develop statistical models of multi-scale sleep EEG dynamics, which will form a framework for identifying biomarkers of psychological and neurological disorders. Sleep is a natural, restorative, altered state of consciousness common to every living human being. Neurophysiologically, sleep is a continuous, dynamic process involving the complex interaction of cortical and sub-cortical networks within the brain operating on multiple time scales. Study of the sleep electroencephalogram (EEG) is therefore an ideal, natural means of simultaneously observing the correlates of activity from numerous brain regions. Fur- thermore, numerous psychological (e.g., schizophrenia, depression, and anxiety) and neurological (e.g., Alz- heimer's disease and Parkinson's disease) disorders are associated with disrupted sleep dynamics, affecting millions of people worldwide. Although ~20 million Americans suffer with disorders of sleep dynamics, such as chronic and/or severe insomnia, current clinical practice does not objectively quantify sleep EEG dynamics. This is because current methods, though instrumental our present understanding of sleep, limit the degree to which sleep dynamics can be described-discretizing the sleep in time and state through subjective, visual scoring practices. Furthermore, the sleep field has yet to adopt powerful spectral estimation techniques, which could greatly improve the characterization of sleep EEG dynamics. Therefore, there would be great benefits in developing objective methods incorporating the state-of-the-art in dynamic modeling and signal processing for sleep dynamics. In order to approach this problem, we have recently developed a signal processing and mod- eling framework to characterize the dynamics of multiple simultaneous neural processes evolving on different time scales. We have shown that these dynamic models significantly outperform traditional stage-based meth- ods in accurately characterizing the behavioral and physiological dynamics of sleep onset. Also, our prelimi- nary studies have shown that optimized spectral estimation methods vastly improve the analysis of sleep EEG dynamics in healthy subjects, and reveal vivid signatures of pathological sleep in Alzheimer's and schizophre- nia patients. In this proposal, we will develop a novel signal processing and dynamic modeling framework to characterize sleep EEG dynamics from a set of large databases (~20,000 records) of sleep recordings from healthy and pathological subjects. By characterizing variability in healthy sleep, we will develop a novel anoma- ly detection approach for identifying and quantifying differences in the sleep EEG dynamics of pathological populations.