An Introduction to Quantitative Analyses of Sleep EEG Via a Wavelet Method

Abstract

Objective: Among various methods developed to quantitatively explore electroencephalograms (EEG), we focused on a wavelet method that was known to yield robust results under nonstationary conditions. The aim of this study was thus to introduce the wavelet method and demonstrate its potential use in clinical sleep studies. Method: This study involved artificial EEG specifically designed to validate the wavelet method. The method was performed to obtain time-dependent spectral power and phase angles of the signal. Synchrony of multichannel EEG was analyzed by an order parameter of the instantaneous phase. The standard methods, such as Fourier transformation and coherence, were also performed and compared with the wavelet method. The method was further validated with clinical EEG and ERP samples available as pilot studies at academic sleep centers. Result: The time-frequency plot and phase synchrony level obtained by the wavelet method clearly showed dynamic changes in the EEG waveforms artificially fabricated. When applied to clinical samples, the method successfully detected changes in spectral power across the sleep onset period and identified differences between the target and background ERP. Conclusion: Our results suggest that the wavelet method could be an alternative and/or complementary tool to the conventional Fourier method in quantifying and identifying EEG and ERP biomarkers robustly, especially when the signals were nonstationary in a short time scale (1-100 seconds).