On the Self-affinity of the Electroencephalogram: Evaluation of a Whole Spectrum of Scale Coefficients by Detrended Fluctuations Analysis

Abstract

Previous studies evaluated the self-affine characteristics of the electroencephalogram (EEG) by Detrended Fluctuations Analysis (DFA). This has been done by computing scaling exponents α over certain time-scales. However, literature is not univocal regarding i) the number of α coefficients which properly describe the EEG self-affinity, and ii) the temporal regions over which the coefficients should be estimated.

We investigated this issue by computing the whole spectrum of scale coefficients, α(n), assessed over a wide range of temporal resolutions, n. This method was then validated by means of synthesized fractal signals. We also assessed the effects on DFA of filters traditionally used for EEG pre-processing.

Results suggest that the self-affine structure of EEG is more complex than usually considered. In particular, EEG self-affinity should be described by a continuous spectrum of scaling exponents, rather than by two or three coefficients only, as done in previous studies. Moreover, filters for signal preprocessing may have dramatic effects on DFA, and could have influenced results previously reported in literature.