Testing procedures for non-stationarity and non-linearity in physiological signals

Area of Science:

• Physiology
• Biophysics
• Complex Systems Analysis

Background:

• Physiological signals (EEG, ECG, blood flow, gait) exhibit complex dynamics, including non-stationarities and non-linearities.
• These signals often resemble red noise with long-range correlations and a 1/(f^beta) power spectrum.
• Classical methods for detecting non-linear dynamics can fail due to short data records and stochastic components.

Purpose of the Study:

• To develop and apply methods for distinguishing underlying signal dynamics from observed time series properties.
• To identify non-stationarities within physiological signals.
• To rigorously test for the existence of non-linear dynamics in these complex signals.

Main Methods:

• Singular spectra analysis and detrended fluctuation analysis were used to assess intrinsic correlation properties and differentiate from external trends.
• Correlation dimension (with a modified re-embedding algorithm) and correlation integrals of real and surrogate data were employed to test for non-linear dynamics.
• The Kolmogorov-Smirnov (K-S) test was used to statistically compare correlation integrals of real signals and surrogate data sets.

Main Results:

• The study successfully applied a suite of methods to analyze physiological signals.
• The techniques were validated on electroencephalogram (EEG) and laser-Doppler (LD) blood flow data.
• Results demonstrated the utility of the combined approach in characterizing complex signal dynamics.

Conclusions:

• No single method is sufficient for analyzing complex physiological signals.
• A comprehensive battery of analytical approaches is recommended for accurately characterizing non-stationarities and non-linear dynamics.
• This integrated methodology enhances the understanding of underlying physiological processes from complex time-series data.