Time-frequency analysis of heart rate variability using SPWD unveiled nonstationary features and idiosyncratic behavior patterns affected by physiological and pharmacological interventions.
Time-frequency analysis using smoothed pseudo-Wigner distribution unveils nonstationary features of heart rate variability dynamics not apparent in standard frequency-domain analysis.
Background: The heart rate variability (HRV) signal is mainly analyzed in frequency‐domain and the signal's spectrum is estimated using either Fast Fourier Transformation (FFT) or the autoregressive (AR) model. These two methods assume that the HRV signal is stationary and additionally the AR method is based on the assumption that the model is linear and the signal is monocomponent in nature. Qualities of spectral estimates are thus closely related to the validity of the above assumptions. Evidence has accumulated indicating that HRV is a multicomponent, nonlinear and nonstationary signal. Then the spectral representations currently used would yield global, approximate, and smoothed view of HRV dynamics. Methods: We applied time‐frequency (TF) analysis methods, smoothed pseudo‐Wigner distribution (SPWD), and spectrogram and complemented for validation by FT spectrum to the HRV signal of fifteen apparently healthy volunteers (mean age 27.2 ± 5.6 years). Short‐term electrocardiograms (ECG) were recorded during supine and upright tilting positions (baseline recording). After baseline recording we induced parasympathetic, sympathetic, and total autonomic blockade correspondingly to six, nine, and four subjects. In addition, in four patients ECGs were recorded during controlled respiration. Results: SPWD and spectrogram revealed strips in frequency, or TF components, corresponding to FT components. High frequency (HF) components appeared stationary (in wide sense), with slight mean frequency shifts during spontaneous respiration, concurrent with respiratory motions. Low frequency (LF) and very low frequency (VLF) components had a nonstationary character displaying activity burst in time and interrelation in frequency. Upright tilting caused a uniform reduction in intensity and bandwidth of the HF component and enhancement of intensity and burst activity of the LF component. There was a pronounced decline of HF and LF components’intensity and decrease of HF component's bandwidth after parasympathetic blockade and total autonomic blockade, while the VLF component did not change. Sympathetic blockade was accompanied by augmentation of the LF and HF components’intensity associated with an increase in the HF component's bandwidth and the spreading of it in the region between the LF and HF. The LF component exhibited less burst activity during tilting under sympathetic blockade, as compared to baseline recordings during tilt. The VLF component's behavior did not change after sympathetic, parasympathetic, and total autonomic blockades. Conclusion: Application of TF distributions to the HRV signal offers a new representation of HRV dynamics. SPWD unveiled features in the HRV signal not available in separate time‐ and frequency‐domains. TF components display idiosyncratic behavior patterns in time and were effected by physiological and pharmacological interventions. A.N.E. 1996;1(4):411–418
Birand et al. (Tue,) conducted a other in Healthy (n=15). Physiological and pharmacological interventions (tilting, autonomic blockades) vs. Baseline recording was evaluated on Heart rate variability dynamics (time-frequency components). Time-frequency analysis of heart rate variability using SPWD unveiled nonstationary features and idiosyncratic behavior patterns affected by physiological and pharmacological interventions.