Correlation Analysis of HRV and Respiratory Signals

Abstract

The variability in human heart rate is regulated by the Autonomic Nervous System (ANS) signal, which can be utilized for monitoring and regulating human stress levels. Additionally, natural breathing produces a cardiac variation called Respiratory Sinus Arrhythmia (RSA). Notably, this variation occurs within the same frequency range as induced by stress, specifically the High Frequency (HF) band (0.12Hz to 0.4 Hz) of Heart Rate Variability (HRV). Multiple techniques can be employed to examine these variations in the human heart, allowing for the assessment of the impact of respiration-induced changes on parasympathetic and sympathetic processes. The technique employed in this research consists of correlational studies between HF- HRV band power and respiration frequency. To conduct this research, two datasets have been taken into consideration. Each dataset consists of 18 subjects. This study is a correlation analysis between heart rate variability (HRV) and respiratory signals to determine the specific frequency range of HRV that selectively accumulates information associated with respiratory sinus arrhythmia (RSA). Hence, it is imperative to examine ECG and Respiratory signals simultaneously. Thus, both datasets contain each subject's simultaneous ECG and respiration signal data. The first dataset is obtained using a custom-designed acquisition system designed for this research, which enables the simultaneous capture of electrocardiogram (ECG) and breathing signals. This system has been developed specifically to possess the qualities of portability, affordability, energy efficiency, and compatibility with USB connections on laptops and PCs. It can deliver performance on par with commercially available equipment of standard quality. The second dataset was acquired from the online data bank MGH/MF database on physionet.org. The HF band of HRV is typically defined as spanning from 0.12Hz to 0.4 Hz, encompassing a broad frequency range that includes most breathing frequencies. An inherent disadvantage of this extensive bandwidth is the possibility of including irrelevant information outside the scope of RSA. A prospective narrower high frequency (HF) band of HRV has been defined to obtain selective respiration-related information, with the HF band centered around the respiratory frequency (RF). This research evaluated the narrower HF-HRV band, which are defined as the respiration frequency (RF) ± a bandwidth (δ). The bandwidth values used in this analysis were 0.04, 0.05, 0.06, 0.07, 0.08, 0.10, and 0.12 Hz. Seven correlational studies have been conducted to determine the narrower band that captures the most information regarding the modulation of heart rate caused by breathing. These correlational studies have been computed between the HF-HRV band's power and respiration frequency. In these seven correlational studies, the value of bandwidth (δ) that is used to define the narrower HF band (𝑅𝐹±𝛿) changes. In the first correlational study, the value of δ was taken as 0.04 Hz. In the subsequent study, δ was taken as 0.05 Hz. This pattern continued with δ values of 0.06 Hz, 0.07 Hz, 0.08 Hz, 0.10 Hz, and 0.12 Hz for the third, fourth, fifth, sixth, and seventh studies. The HF-HRV band power has been calculated using the Fast Fourier Transform (FFT) for all the subjects in both datasets over all the narrower bands. The study's findings indicate that the mean average negative correlation has been calculated by averaging the average correlation values obtained from seven correlational studies in both datasets. These mean average correlation for the HF bands, denoted as (RF±0.04), (RF±0.04), (RF±0.04), (RF±0.04), (RF±0.04), (RF±0.04), is calculated to be -0.3999, -0.4175, -0.4478, -0.4525, -0.4528, -0.4593, and -0.4506 respectively. Thus, the narrower HF band, defined by (RF±0.10), effectively collects the maximum information about the variation of heart rate generated by breathing (or respiration). The second objective of this study is to determine the optimal computational technique for computing the selected narrower HF-HRV band (RF±0.10) power to achieve the maximum negative correlation with respiration frequency. The computational methods that have been explored include the Fast Fourier Transform (FFT) applied to the entire Heart Rate Variability (HRV) signal, the Hilbert-Huang Transform (HHT) applied to the entire HRV signal, the FFT method applied to the first Intrinsic Mode Function (IMF) of the HRV signal, and the HHT method applied to the first IMF of the HRV signal. This study computes the mean average correlation by averaging the correlation values acquired from four computing processes to determine the most effective computational technique. The mean average correlation values for the computational techniques are as follows: -0.4267 for the FFT method on the entire HRV signal, -0.3924 for the HHT method on the entire HRV signal, -0.4748 for the FFT method on the first IMF of the HRV signal, and -0.4287 for the HHT method on the first IMF of the HRV signal. The results indicate that the average negative correlation obtained from the FFT technique applied to the first Intrinsic Mode Function (IMF) of the HRV signal, employing a narrower HF band is the most robust among all other correlations. Hence, employing this computational technique to selectively capture information related to RSA is advisable. This HF-HRV band power can be utilized to differentiate between Normal Sinus Arrhythmia (NSA) and non-NSA, in addition to monitoring and regulating human stress.