Background: Over the last two decades the advances in analysis techniques for physiological time series data have been moving from the classical statistics to a more nonlinear or chaos based approach to looking at patterns in the variability of the time series. From this work it can be shown that physiological time series exhibit complex multi-fractal properties. So by designing a classification based on this nonlinear and chaotic nature you can detect changes and alterations in the underlying physiological processes.
Methods: Applying a proven relationship between the wavelet modulus maxima representation and the Hölder exponent we could assess the multi fractal nature of the of the signal detection underlying changes in the physiology. Using two distinct techniques one global and the other localised in time, classification of two distinct the time series was carried out firstly via the analysis of the distribution of the Hölder exponents over all scales of the signal and secondly via a moving window application of the mean Hölder function.
Findings: The distribution methodology did not return significant results though this is probably more to do with the signal than the technique. The trending approach shows a predictive nature with slope being linked to increased instability in the signal content.
Conclusions: Overall this study has shown the applicability of the techniques which definitely warrant further refinement and study.