Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 186

(Sect. 2.4.1). In Figure 4.34, the first four components, difference between compo-
nent one and four, and Hilbert amplitude ratio of third and first component (a3/a1) as
a function of time are illustrated. The value of component ratio may be considered as
a convenient measure of the HRV rhythms evolution. One can observe the increase
of the ratio a3/a1 after changing posture.
FIGURE 4.35: Instantaneous frequency as a function of time for the first four
components presented in Figure 4.34. From [Echeverria et al., 2001].
In Figure 4.35 instantaneous frequency as a function of time for first components
presented in Figure 4.34 is shown. The instantaneous frequency (IF) seems to be
a good index for tracing the changes in HRV rhythms, but in the procedure of IF
determination appropriate boundary requirements have to applied, which is not a
trivial problem [Huang et al., 1998]. Another problem in application of EMD is the
reproducibility of results. In the study of [Maestri et al., 2007] (described in more
detail in Sect. 4.2.2.4.5) the reproducibility of EMD was found to be very poor. It
seems that the EMD method has to be used with caution, especially in respect of
fulfilling the assumptions of the method and in respect of statistical significance and
reproducibility of results.
4.2.2.4.2 Entropy measures The indexes connected with information theory
(Sect. 2.5.6): approximate entropy (ApEn) [Pinkus and Singer, 1996], conditional
entropy [Porta et al., 1999], and sample entropy (SaEn) [Richman and Moorman,
2000] were introduced and found applications in recognition of certain cardiovascu-
lar pathologies. For example in [Perkiomaki et al., 2005] a progressive decrease in
ApEn indicating reduced complexity of heart rate variability was observed before the
onset of atrial fibrillation. Conditional entropy was used for evaluation of regularity,