Abstract
Symbolic dynamic analysis (SDA) methods have been applied to biomedical signals and have been proven efficient in characterising differences in the electroencephalogram (EEG) in various conditions (e.g., epilepsy, Alzheimer’s, and Parkinson’s diseases). In this study, we investigated the use of SDA on EEGs recorded during sleep. Lempel-Ziv complexity (LZC), permutation entropy (PE), and permutation Lempel-Ziv complexity (PLZC), as well as power spectral analysis based on the fast Fourier transform (FFT), were applied to 8-h sleep EEG recordings in healthy men (n=31) and women (n=29), aged 20-74 years. The results of the SDA methods and FFT analysis were compared and the effects of age and sex were investigated. Surrogate data were used to determine whether the findings with SDA methods truly reflected changes in nonlinear dynamics of the EEG and not merely changes in the power spectrum. The surrogate data analysis showed that LZC merely reflected spectral changes in EEG activity, whereas PE and PLZC reflected genuine changes in the nonlinear dynamics of the EEG. All three SDA techniques distinguished the vigilance states (i.e., wakefulness, REM sleep, NREM sleep, and its sub-stages: stage 1, stage 2, and slow wave sleep). Complexity of the sleep EEG increased with ageing. Sex on the other hand did not affect the complexity values assessed with any of these three SDA methods, even though FFT detected sex differences. This study shows that SDA provides additional insights into the dynamics of sleep EEG and how it is affected by ageing.
Highlights
Neuronal interactions in the brain are highly nonlinear, making nonlinear time series analysis methods appropriate for the characterisation of electroencephalogram (EEG) recordings [1]
Lempel-Ziv complexity (LZC), permutation entropy (PE), and permutation Lempel-Ziv complexity (PLZC) were computed on each sleep epoch (30-s) for each vigilance states (VS)
PE and PLZC yielded similar results. Both methods could distinguish between brain activities in different VS and detect physiological changes that affect the structural mechanisms underlying the brain activity, in agreement with previous studies [52]. These results show the potential usefulness of these methods in sleep research, where nonlinear dynamic changes in the brain activity could be characterised by relative increases and decreases in complexity
Summary
Neuronal interactions in the brain are highly nonlinear, making nonlinear time series analysis methods appropriate for the characterisation of electroencephalogram (EEG) recordings [1]. The introduction of nonlinear methods for the analysis of EEG signals [2, 3] made the study of complex neural networks in the brain possible in ways that were not feasible with linear methods, such as the Fourier transform (FT). Many of the classic methods used to this aim can provide spurious results when computed from noisy time series, such as EEG recordings [4]. Research into nonlinear methods that are well-suited to the analysis of noisy complex time series, such as the EEG, is required. The computational efficiency is better than that of other nonlinear
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
