Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 187

synchronization, and coordination in cardiovascular beat-to-beat variability [Porta
et al., 2000].
SaEn was applied for a study of heart rate variability during obstructive sleep
apnea episodes by [Al-Angari and Sahakian, 2007]. When compared with spectral
analysis in a minute-by-minute classification, sample entropy had an accuracy, sensi-
tivity and specificity slightly worse than the results supplied by the spectral analysis.
The combination of the two methods improved the results, but the improvement was
not substantial. Nevertheless, SaEn has begun to be applied in practice, namely in
wearable devices. It was reported that SaEn was used to assess respiratory biofeed-
back effect from HRV [Liu et al., 2010].

4.2.2.4.3 Detrended fluctuation analysis Another measure of complexity is pro-
vided by detrended fluctuation analysis (DFA). In its framework the fractal index of
self-similarity is determined (Sect. 2.5.3).

FIGURE 4.36: Scaling properties of 24 hour tachogram of a healthy subject. RR
interval tachogram (upper left), power spectra (lower left), a power-law scaling slope
of long-term fluctuations of heartbeats (upper right), and DFA results (lower right).
β—long-term scaling slope, α 1 short-term fractal scaling exponent (4-11 beats), α 2
intermediate scaling exponent ( > 11 bits). From [Perkiomaki et al., 2005].

The HRV scaling exponent value depends on the window length. In Figure 4.36 the
examples of DFA results illustrating dependence of scaling exponents on data length
are shown. In the lower right part of the figure it may be observed that the slope
of the log-log plot changes with increase of the window size > 11 heartbeats. The