Multifractal analysis of electroencephalogram zero-crossing time intervals

Ma Qian-Li,Bian Chun-Hua,Wang Jun,Ning Xin-Bao

DOI: https://doi.org/10.3321/j.issn:0469-5097.2008.04.002

2008-01-01

Abstract:Electroencephalogram(EEG) is a typical signal which is particularly susceptible to biological and experimental artifacts that manifest themselves as large amplitude deviations,and this leads to a questioning of the validity of techniques based on generalizations of geometric box counting,such as various approaches of the multifractal analysis.Therefore,zero-crossing time intervals(CTI) which were derived from zero-crossings of EEG time series were introduced as study subjects.Zero-crossings is a frequency-based alternative to the amplitude millivolt fluctuations currently used as the main time series for EEG analysis,which might be more robust in light of fluctuations due to amplitude artifacts.A new approach for studying nonlinear dynamics of EEG by applying multifractal analysis to CTI series of EEG was proposed.The EEG data used in this study come from PhysioBank "The Sleep-EDF Database".We applied multifractal analysis to CTI series of sleep EEG,and the result confirms multifractality in EEG CTI series.We studied 24h EEG records of four subjects(sc4002e0,sc4012e0,sc4102e0 and sc4112e0).The signals were divided into segments of 30 seconds in order to be consistent with the hypnograms.Singularity strength range Δα was estimated for quantifying changes of multifractal singularity spectra.We observed that Δα increased when the brain went deeper sleep.In comparison,both multifractal analysis and DFA were performed on CTI and amplitude fluctuation time series of EEG.We found that ΔαCTI and Δαamp both increased as sleep went deeper.Yet DFA scaling exponents α of CTI and amplitude fluctuations varied contrarily with sleep state transition: αCTI decreased as sleep went deeper,while αamp increased.We did ANOVA followed by multiple comparison by using Tukey-Kramer method to test the significance of the difference among the stages.And ANOVA gave a p-value of P<0.001 for all the four measures,suggesting significant differences between the means of stages,and ΔαCTI gave the best discrimination according to multiple comparison test since only one pair of the means was detected with no significant difference.The evidence of multifractal structures in EEG CTI series was found,suggesting multifractality of EEG which coincides with that of previous studies based on multifractal analysis of amplitude fluctuations.And the range of singularity strength increased when the brain went deeper sleep,indicating that the scaling properties of EEG becomes more inhomogeneous in deeper sleep.The test results of 24h sleep EEG study suggest that multifractals are more appropriate for characterizing nonlinear dynamics of EEG than monofractals,and CTI are better series than amplitude fluctuations for studying fractal properties of EEG.As a conclusion,it is suggested that multifractal analysis of EEG CTI series seems suitable for automatic discrimination of sleep stages,and could also be fit for nonlinear dynamics studies of brain functions and pathological EEGs.