睡眠脑电图多种分析方法的差异比较

Difference of multiple analytical methods of electroencepha-logram in sleep

目的:探讨睡眠脑电图分析多种方法的特征及其差异。资料来源:应用计算机检索Medline1985-01/2004-12与睡眠脑电相关的文献,检索词“EEG,sleep,nonlinear-analysis”,并限定文献语言种类为英文。同时检索万方数据库1995-01/2004-08与睡眠脑电相关的文献,检索词“睡眠,脑电,非线性分析”,并限定文献检索语言种类为中文。资料选择:从资料中选取与睡眠脑电研究方法相关的文献。纳入标准:①脑电信号处理的传统方法。②脑电信号处理的现代方法。排除标准:重复研究、综述类文章。资料提炼:共收集到45篇关于睡眠脑电研究方法的文章,15篇符合纳入标准。排除的30篇都是重复的同一研究。15篇文章分别用不同的方法对脑电信号进行研究,是各方法有代表性的文章。资料综合:①经典分析方法主要是由专家对连续记录的整夜睡眠图形数据进行目测分析后,对睡眠过程中不同期间的睡眠深度进行评估。②人工神经网络分析方法最主要的特征为连续时间非线性动力学、网络的全局作用、大规模并行分布处理及高度鲁棒性和学习记忆功能。③维数分析数值结果只能显示各状态间的比较差异,这在某种程度上使其受到限制。④近似熵是用一种有效的统计方式,即边缘概率的分布来区分各种过程,是测量序列的复杂性的一种方法,能更多地提取出序列的复杂性信息。⑤小波分析可以由粗及细地逐步观察信号,适当地选择基本小波,可以使变换在时、频两域都具有表征信号局部特征的能力,实现睡眠分期的检测。结论:睡眠是一种复杂的生理过程,研究脑电可以从本质上对睡眠进行分析。由于脑电自身的复杂性,引入非线性分析方法,使得在睡眠脑电的研究上取得进步。这些现代分析方法将为睡眠的监测和质量的研究做出更大的贡献。

OBJECTIVE： To investigate the characteristics and the differences of the multiple methods for the analysis of electrocncephalogram in sleep. DATA SOURCES： An online search of Medline was undertaken to identify English articles related to electrocncephalogram in sleep published between January 1985 and December 2004 by using the keywords of “electroenceph alogTam, sleep, nonlinear-analysis”. Articles about electroencephalogram in sleep published in Chinese from January 1995 to August 2004 were searched with the keywords of “sleep, electroencephalogram, nonlinear-analysis” in Wanwang database. STUDY SELECTION： The articles about the investigative methods of the electroencephalogram in sleep were selected. Inclusion criteria： ① traditional methods for electroencephalogram signal processing; ②Modern methods for electroencephalogram signal processing. Exclusion criteria： repetitive studies and reviews. DATA EXTRACTION： Totally 45 articles about the investigative methods of electroencephalogram in sleep were collected, and 15 of them were in accordance with the inclusion criteria, all the 30 being excluded ones were repetitive studies. The 15 articles studied electrocncephalogram signals with different methods respectively, they were all representative. DATA SYNTHESIS： ① Classic analytical methods mainly evaluated the sleeping depth at different period during sleep after the continuously recorded all-night sleep figure and data were measured with eyes and analyzed by the specialists. ② The most capital characteristic of the artificial neural network analytical method was the continuous time-related nonlinear dynamics, the overall role of network, paralleled distributive treatment with large scale, high robust and learning and memory function. ③ The results of dimension analysis could only show the comparative differences among the status, which was restricted to some degree. ④ Approximate entropy was an effective statistical method, that was, to differentiate each process according to the distribution of marginal probability, and it was a method to measure the complicity of sequence, and it can extract more information about the complicity of sequence. ⑤ Wavelet analysis could observe signals step by step from thick to thin; Suitable selection of basic wavelet could make the transformation have the ability in manifesting the local characteristics of signals in both fields of time and frequency, so as to realize the staging detection of sleep. CONCLUSION： Sleep is a complex physiological process, and the study of electroencephalogram can analyze sleep essentially. Because of the complexity of electrocncephalogram itself, the introduction of nonlinear analysis makes progress for the study of electroencephalogram in sleep. These modem analytical methods will contribute more for the studies about the monitoring and quality of sleep.