睡眠脑电的非线性动力学方法

EEG time-series analysis using nonlinear dynamics method for sleep monitoring

在8例健康成年人的睡眠脑电监测实验基础上,利用已有的专家人工分期结果,提取睡眠各阶段特征数据,应用近似熵、复杂度和功率谱熵三种方法进行分析,从客观量化的复杂性度量来刻划睡眠深度的变化情况.对每个睡眠分期选取5 000点数据,数据窗取1 000点,逐次延时一个采样间隔得到几个时间序列,分别求复杂度,最后取均值即得此分期复杂性测度值.结果表明三种方法均与专家人工分期结果相吻合.近似熵算法复杂不适合在线分析;复杂度算法较简单,但数据粗粒化处理容易丢失信息;功率谱熵算法简单、快速及有效.因而用统计分析方法分析,表明功率谱熵能较好地反映睡眠深度的变化情况.

Approximate entropy, Lem-Ziv complexity and power spectral entropy (PSE) are nonlinear dynamic methods to measure Electroencephalograph (EEG) time-series complexity in recent years. EEG is a kind of biomedical signal to monitor the depth of sleep. The algorithm of approximate entropy, Lem-Ziv complexity and power spectral entropy are introduced. A discussion is made on their merits and demerits. For 8 healthy volunteers without any medication, analysis of EEG using above mentioned methods is performed. Results show that the PSE of EEG signals during sleep process can correctly affect sleep deepness. and accord with the results of sleep stage by expert.