Application of wavelet transform and wavelet entropy in the characteristic study of sleep electroencephalogram signal change
Yu-liang Xiao,Wei-xing He,Xiao-ping Chen,Yi Ji,Qiu-ming Wu
DOI: https://doi.org/10.3321/j.issn:1673-8225.2006.25.047
2006-01-01
Abstract:Aim: To analyze and compare the dynamic features and statistical characteristics of the complexity measure for sleep electroencephalogram (EEG) in each vigilant state with wavelet entropy (WE). Methods: This experiment was carried out at the laboratory of bioengineering department of College of Electronic and Information Engineering of Jiangsu University from August 2005 to September 2005. Experimental data were obtained from MIT-BIH database. One channel overnight sleep EEG signal was adopted from one of eight testees. No sleeping illness was found in each testee. Then, we analyzed and studied the complexity of sleep EEG based on experiment sleep data, the wavelet packet de-noising EEG signals were decomposed into four components of δ, θ, α, and β by using orthogonal wavelet transform (WT). Then, the wavelet entropy of four components and the summation were calculated as a function of time. Results: Experimental results showed that there were significant differences among the WEs of EEGs recorded in the six states of sleep stage. There was the maximal WE value in the resting conscious stage. Among the four sleep stage of non-rapid eye movements (NREM), WE value was the maximal in S1 sleep stage and the minimal in S4 sleep stage, WE value decreased in turn from slight sleep to deep sleep. WE value of rapid eye movement (REM) sleep stage was between resting conscious stage and NREM sleep stage. Among the four basal rhythm bands (δ, θ, α and β) of EEG, the common characteristic was that WE value was minimal in the β band and WE value was maximal in the θ band. In the δ, α and β bands, WE value of resting conscious stage was maximal, WE value decreased in turns from S1 sleep stage to S4 sleep stage, WE value of REM sleep stage was between resting conscious stage and NREM sleep stage. The fluctuated range of WE value of the four basal rhythm bands was obviously larger during S1, S2, S3 and S4 sleep stage than resting conscious stage and NREM sleep stage. Variance value tendency was similar to mean value and more obvious in each sleep stage and frequency band. Conclusion: Experimental results show that WE value used to analyze the complexity measure for sleep EEG decreased in turns from slight sleep to deep sleep, which is consistent with theoretics. WE can be used as a detected index of change characteristics of EEG under different physiological status. WE can be used not only to distinguish the long-term changes in EEG complexity, but also to reveal the short-term changes in EEG microstate.