Effects of electrocardiographic noise on ultra-short term Heart Rate Variability indices

Abstract

Heart Rate Variability (HRV) is a key metric for assessing cardiovascular health and autonomic nervous system function. The increasing use of wearable devices for continuous health monitoring during daily-life activities presents significant challenges, since the acquired signals are often noisy or affected by artifacts, resulting in a low signal-to-noise ratio (SNR). This study aims to investigate how electrocardiographic (ECG) noise affects the accuracy of ultra-short term (∼ 2 min) HRV analysis. Time-, frequency- and information-domain HRV indices, computed on interbeat interval time series extracted from ECG signals contaminated by different types of simulated noise (white and frequency-specific) at various SNR levels (−3, 1, 5, 10 and 20 dB) were compared to those obtained on reference noise-free waveforms. The results show that low-frequency noise (i.e., at 0.01, 0.1, 0.3, and 3 Hz) at an SNR lower or equal than 5 dB has a significant impact on the reliability of HRV measures, leading to remarkably diminished correlation with reference values. On the other hand, white and higher-frequency noise (i.e., 50 Hz and 300 Hz) had a reduced impact on the computed indices even for very low SNR values. Overall, a SNR level of at least 10 dB seems enough for ensuring reliable HRV analysis across all domains. These findings are valuable for improving the reliability of HRV analysis especially in the case of short-duration signals acquired in noisy or extreme environments, to ensure that wearable devices can provide reliable physiological information even in challenging conditions.