THE VALIDITY OF RHRV IN BATCH PROCESSING RAW EKG SIGNALS FOR THE DETERMINATION OF HEART RATE VARIABILITY
Travis Anderson, Jennifer L. Etnier, Emily E. Bechke, Laurie Wideman, FACSM. University of North Carolina at Greensboro, Greensboro, NC.
BACKGROUND: Heart rate variability (HRV) metrics are commonly utilized in various exercise settings and health sciences. In determining HRV, raw electrocardiograph (EKG) signals are typically individually and manually analyzed, greatly increasing the time burden and the risk of inter- and intra-rater discrepancies. Analyzing raw EKG signals via replicable computer code to determine HRV would address these issues and allow for comparability across studies. METHODS: Forty-four raw EKG signals collected as part of a larger study were used to test the validity of a custom R script for analyzing raw EKG signals. The script utilized the RHRV package and compared it to criterion HRV values obtained from Kubios Premium (v3.1.0). Calculated HRV metrics (standard deviation of the normal-normal intervals [SDNN], root mean square of successive differences [RMSSD], and high-frequency power [HF]) were compared to Kubios values. Comparisons were made under three conditions to iteratively and progressively test for analytical differences between Kubios and the custom R code results: first, by using the unfiltered RR time series obtained from Kubios; secondly, by using Kubios automatic artifact-corrected values; and finally, by using the custom and adjustable R code R-wave detection algorithm and artifact corrections. Raw EKG signals were collected at either 1000 or 2000 Hz, and epoch (range: 90 to 300 seconds) for analysis was selected in Kubios. Custom R-wave detection in R was completed via a version of the Pan-Tomkins algorithm, with varying and optimized parameters for the Butterworth bandpass filter. Custom R code artifact correction was completed via an adaptative threshold algorithm. The validity of calculated HRV metrics was assessed via validity coefficients. RESULTS: Uncorrected signals demonstrated high validity between methods for SDNN (r = 0.9999), RMSSD (r = 0.9999), and HF (r = 0.9219). The progressive addition of artifact corrections (SDNN: [r = 0.9869]; RMSSD [r = 0.8900]; HF [r = 0.7349]) and R-wave detection (SDNN: [r = 0.7691]; RMSSD [r = 0.5401]; HF [r = 0.6120]), reduced the agreement between methods. CONCLUSIONS: These results suggest that HRV can be reliably and validly calculated using RHRV, but methodologic differences in artifact correction and R-wave detection commonly completed for all HRV analysis can substantially affect the calculated HRV metrics.
Anderson, T; Etnier, JL; Bechke, EE; and Wideman, FACSM, L
"THE VALIDITY OF RHRV IN BATCH PROCESSING RAW EKG SIGNALS FOR THE DETERMINATION OF HEART RATE VARIABILITY,"
International Journal of Exercise Science: Conference Proceedings: Vol. 16:
1, Article 210.
Available at: https://digitalcommons.wku.edu/ijesab/vol16/iss1/210