Coherence Estimation of Heart Rate Variability and Respiratory Signals

Spectral analysis of Heart Rate Variability (HRV) has become increasingly common for non-invasive assessment of autonomic cardiac regulation. The HRV signal is extracted from the ElectroCardioGraphic (ECG) signal from the heart and is normally sampled with 1000 Hz. The so-called RR-intervals, which are the distances between two following heart beats, are detected and each interval length is represented as a level of a length corresponding to the interval length. The created HRV-signal is then a staircase signal which is downsampled to 4 Hz, generating a signal to be analyzed for the variation off the heart rate. Within the HRV power spectrum, three frequency components, which are suggested to reflect different neurally mediated oscillations, are commonly of interest. A region at around 0.03 Hz, the very low frequency band (VLF), is located and around 0.1 Hz, the so-called low frequency band (LF). Of interest here is the power of the high frequency component (HF-HRV) that is negatively related to respiration, i.e., respiratory sinus arrhythmia (RSA), which mirrors parasympathetic or vagal regulation of the heart. In view of the substantial amount of research relating psychosocial stressors in work life to cardiovascular disease, it is important to reliably estimate the power of the HF-HRV. The aim of this project is to investigate different spectral techniques to estimate and follow changes of the power of the HF-HRV component. An example of the analysis is depicted where frightening pictures have been shown in an experiment every 30 s in a 5 minutes long, where the intention was to examine if the effect in RSA-magnitude would vary as a function of stimulus interval (i.e. every 30 s). The multitaper spectrogram using the Peak Matched Multiple Windows (PM MW) is shown below, where the changes in frequency and increase in amplitude (red color) is clearly visible with a period of 30s. We have also suggested a narrower high frequency-band (HF) based on the respiratory peak frequency for the power estimation of the HF-band which has been compared with power estimation using the usual HF-band (0.12-0.4Hz). The results show a significant improvement in the robust estimation of HF-HRV power.