TABLE OF CONTENTS It should be noted at this point that all of the traditional HRV indices employ techniques that assume (weak) stationarity in the data. If part of the data in the window of analysis exhibits significant changes in the mean or variance over the length of the window, the HRV estimation technique can no longer be trusted. A cursory analysis of any real RR tachogram reveals that shifts in the mean or variance are a frequent occurrence [94]. For this reason it is common practice to detrend the signal by removing the linear or parabolic baseline trend from the window prior to calculating a metric.
However, this detrending does not remove any changes in variance over a stationarity change, nor any changes in the spectral distribution of component frequencies. It is not only illogical to attempt to calculate a metric that assumes stationarity over the window of interest in such circumstances, it is unclear what the meaning of a metric taken over segments of differing autonomic tone could be. Moreover, changes in stationarity of RR tachograms are often joined by transient sections of heart rate overshoot and an accompanying increased probability of artifact on the ECG (and hence missing data) [86, 95].
In this section we will explore a selection of methods for dealing with nonstationarities, including multiscale techniques, detrending, segmentation (both statistically and from a clinical biological perspective), and the analysis of change points themselves.
Empirical analyses employing...
