基于条件概率的短时睡眠状态实时估计

目的根据脑电信号的特征,提出基于条件概率的睡眠状态实时估计方法,为睡眠监测提供反映睡眠状态连续变化的客观评价依据。方法在白天短时睡眠过程中,同步采集了4导与睡眠相关的脑电信号(C3-A2,C4-A1,O1-A2,O2-A1),对每5秒记录数据进行傅里叶变换,分别计算了8~13 Hz和2~7 Hz的脑电节律能量占空比特征参数。主要方法包含了学习和测试两个阶段:在学习阶段,根据训练数据获得脑电特征参数的概率密度分布;在测试阶段,根据当前特征,得到各睡眠分期的条件概率,并计算获得睡眠状态的估计值。结果分析和测试了12名受试者的短时睡眠数据。通过与睡眠分期的人工判读结果相比较,睡眠状态估计值呈现了睡眠深度的连续变化。觉醒期的显著性差异为2.94,睡眠一期和二期分别为1.78和1.62,分析结果符合实际规律。结论本文所定义的睡眠状态估计值蕴含了睡眠分期的特征,较好地反映了睡眠阶段在持续和过渡期间的连续变化过程,能够为白天短时睡眠状态分析提供实时监测和分析的客观评价依据。

Sleep level estimation based on conditional probability for nap

Objective According to the characteristics of electroencephalograph( EEG),an automatic sleep level estimation method based on conditional probability is developed. The ultimate purpose is to obtain and realize the real-time sleep level evaluation. Methods There are 4 EEG channels(O2 -A1 ,O1 -A2 ,C4 -A1 , C3 -A2 )recorded during nap. For every 5-second data,two characteristic parameters of ratio of EEG rhythms (8-13 Hz,2-7 Hz)are calculated after fast Fourier transformation(FFT). The main method consists of two models:learning and testing. During the learning stage,the probability density functions of EEG parameters are obtained based on the training data. During the testing stage,the sleep level is estimated based on the conditional probability of sleep stages. Results The nap data of 12 subjects are tested. The significant difference is analyzed. Comparing with the visual inspection of sleep stage,the obtained sleep level reflects the continuous change within or between sleep stages. The significant difference of stage awake is 2. 94,stage 1 is 1. 78 and stage 2 is 1. 62,which fits to the regular patterns. Conclusions The defined sleep level is effective to observe the changes within the sleep stage and the transition process between the sleep stages. This method is usable for real-time nap and sleep level evaluation.