Characterization of Dynamic Interactions Between Cardiovascular Signals by Time-Frequency Coherence

Abstract

An assessment of the dynamic interactions between cardiovascular signals can provide valuable information to improve the understanding of cardiovascular control. In this study, two methodologies for the characterization of time-frequency (TF) coherence between cardiovascular signals are described. The methodologies are based on the smoothed pseudo-Wigner-Ville distribution (SPWVD) and multitaper spectrogram (MTSP), and include the automatic assessment of the significance level of coherence estimates. The capability to correctly localize TF regions, where signals are locally coupled, is assessed using computer-generated data, and data from healthy volunteers. The SPWVD allows for the localization of these regions with higher accuracy (AC > 96.9% for SNR ≥ 5 dB) than the MTSP (AC > 84.4% for SNR ≥ 5 dB). In 14 healthy subjects, TF coherence analysis was used to describe the changes, which a tilt table test provokes in the cardiovascular control. Orthostatic stress provoked an increase in the coupling between R-R variability (RRV) and systolic arterial pressure variability; it did not provoke any significant changes in the coupling between RRV and respiration. In HF band, it decreased the strength of the coupling between RRV and pulse interval variability estimated from arterial pressure signal.

A Time-Varying Nonparametric Methodology for Assessing Changes in QT Variability Unrelated to Heart Rate Variability

“…Spectral analysis is extended to the TF domain by means of the quadratic Cohen's class distributions as described in previous studies [27], [28]. Briefly, auto and cross-spectra of two non-stationary signals x(t) and y(t) are estimated as:…”

“…where xy (τ, ν) is the narrow-band symmetric ambiguity function of signals x(t) and y(t), and φ d-D (τ, ν) is an elliptical exponential kernel defined in the ambiguity function domain as [27]:…”

“…The low-pass filtering performed by the kernel in (8) is necessary to remove the interference terms that characterize all quadratic TF distributions and the Wigner-Ville distribution in particular [27]. The kernel function determines the degree of filtering and interference-terms reduction, which is particularly important in coherence analysis since reliable coherence estimates are obtained only if interference terms are completely removed [27].…”

When Interpolation-Induced Reflection Artifact Meets Time–Frequency Analysis

“…The adaptive harmonic model has been applied to study the "non-stationary" physiological dynamics [6,15,28,29,23,16]. Before proceeding, we comment that most time series we could acquire in the real world are real, so we consider only real model but not the exponential complex model.…”

“…There have been several TF analysis techniques proposed, including linear methods like short time Fourier transform (STFT), continuous wavelet transform (CWT) [7,8] and multiwindow approach [9], the quadratic methods like Wigner-Ville distribution and Cohen class [8], nonlinear methods like reassignment (RM) technique [10,11], synchrosqueezing transform (SST) [12,13], multi-tapered RM [14], multi-tapered SST [15,16], ConceFT [17], empirical model decomposition [18], sparse TF analysis [19], iterative filtering [20], etc. The potential of the TF analysis has been shown in several different fields, in particular the medical field; for example, [21,22,23,15], to name but a few.…”