Methodological considerations in calculating heart rate variability based on wearable device heart rate samples

Area of Science:

• Cardiovascular physiology
• Biomedical signal processing
• Wearable technology

Background:

• Heart rate variability (HRV) analysis is increasingly used in wearable devices.
• HRV relies on heart beat intervals, often non-uniformly sampled.
• Spectral HRV analysis typically requires uniformly resampled data (e.g., 4 Hz).

Purpose of the Study:

• To validate the use of 1 Hz resampled data from wearable devices for HRV analysis.
• To compare spectral HRV results obtained at 1 Hz and 4 Hz resampling rates.
• To assess the accuracy of 1 Hz data in representing wearable device BPM time series for HRV.

Main Methods:

• Comparison of spectral HRV parameters (LFnu, HFnu, LFnu/HFnu) at 1 Hz and 4 Hz resampling rates.
• Application of specific signal processing techniques to non-uniformly sampled RRI data.
• Calculation of mean relative errors between the two resampling rates.

Main Results:

• Lowest mean relative errors for LFnu, HFnu, and LFnu/HFnu between 1 Hz and 4 Hz were 3.7%, 15.3%, and 16.4%, respectively.
• 1 Hz resampling provides a reasonable estimation of sympathetic activity.
• 1 Hz resampling results in a poor estimation of parasympathetic activity.

Conclusions:

• 1 Hz sampled heart rate data from wearables can be used for HRV analysis with acceptable sympathetic activity estimation.
• Accuracy of parasympathetic activity estimation using 1 Hz data is limited.
• Signal processing techniques can mitigate some errors but do not fully resolve parasympathetic assessment limitations.