Dynamical heart beat correlations during running

Abstract

fluctuations of the human heart beat constitute a complex system that has been studied mostly under resting conditions using conventional time series analysis methods. During physical exercise, the variability of the fluctuations is reduced, and the time series of beat-to-beat RR intervals (RRIs) become highly non-stationary. Here we develop a dynamical approach to analyze the time evolution of RRI correlations in running across various training and racing events under real-world conditions. In particular, we introduce dynamical detrended fluctuation analysis and dynamical partial autocorrelation functions, which are able to detect real-time changes in the scaling and correlations of the RRis as functions of the scale and the lag. We relate these changes to the exercise intensity quantified by the heart rate (HR). Beyond subject-specific HR thresholds the RRIs show multiscale anticorrelations with both universal and individual scale-dependent structure that is potentially affected by the stride frequency. These preliminary results are encouraging for future applications of the dynamical statistical analysis in exercise physiology and cardiology, and the presented methodology is also applicable across various disciplines. The increasing popularity and accuracy of wearable devices and sensors present new opportunities to study human physiology in a continuous, non-invasive manner for a huge number of subjects under real-world conditions. These devices enable the measurement of a plethora of physiological and mechanical signals such as the heart rate, beat-to-beat (RR) intervals, overall motion via GPS, motion of specific body locations via accelerations, and skin temperature. These data can be recorded in real time, often at 1 s intervals, and uploaded to web services. To date, most recorded data are not analyzed in scientific rigour due to a lack of suitable models for the dynamics of physiological signals under various intensities of exercise load, and also due to restricted availability of the data (property of industry and users). This limits opportunities for a better understanding of complex physiological processes, diagnostics and monitoring for patients in rehabilitation, and the optimal training of athletes. However, it has been long known that a variety of physiological conditions and cardiac diseases affect heart rate variability (HRV) and the correlations in RR intervals 1. In exercise physiology, HRV is often used at rest to evaluate recovery, fatigue and overtraining. It is known that during exercise the overall variability of the RR intervals (RRI) is strongly suppressed. Regardless, the RRI correlations contain valuable information even during exercise 2-4. For example, the possibility to determine certain physiological thresholds, such as the anaerobic threshold, from the frequency spectrum of HRV has been examined 5,6. Often the relative importance of lowfrequency (LF: 0.04-0.15 Hz) and high-frequency (HF: 0.15-0.4 Hz) spectral power is studied during exercise. Using this concept as a measure of t...

Online Tool for Dynamical Heart Rate Variability Analysis

“…The most glaring difference is the lack of a correlated band at the shortest scales. The appearance of short-scale anticorrelations during high intensity exercise and their broadening with increasing heart rate appears to be universal, but the details of the correlation landscapes vary, which is also consistent with the results during running [5].…”

“…The DDFA scaling exponent α(t, s) is computed in segments of length l DDFA s. Hence, this parameter controls the balance between locality and statistical noise in the DDFA results. Generally the choice l DDFA = 5 is suitable for HRV analysis [5].…”

“…Dynamic smoothness priors detrending is an option for explicit detrending of the RRI data before the analysis. While the DFA method performs intrinsic detrending during the analysis, it is limited to polynomial trends and higher order detrending has issues of its own [5]. Therefore, we employ the conventional smoothness priors detrending [8] by separately detrending the data for each scale s before DDFA.…”

“…In some situations, such as during sports, it is desirable to study transient changes in HRV. To this end, a dynamical approach to DFA has been developed, which has been successfully employed in studying RRI correlations during running and different sleep stages [5,6]. Here we present an online tool for dynamical HRV analysis that implements the method.…”

Correlation Properties of Heart Rate Variability during a Marathon Race in Recreational Runners: Potential Biomarker of Complex Regulation during Endurance Exercise

“…Despite a reduction of this external load indicator, the DFA-alpha1 continued to fall to values seen in the highest intensity domain during incremental exercise testing in agreement with lactate assessment. Therefore, the discrepancy between the reduced running pace with that of the decline of DFA-alpha1, may demonstrate that there is a complex modulatory dynamics present during the course of the race that may not be detected by conventional linear HRV measurements (Molkkari et al, 2020). Therefore, the dimensionless biomarker DFA-alpha1 may demarcate the complex dynamics of internal load development over the course of a marathon race as a descriptor of NPE (Balagué et al, 2020).…”