Heart Rate Variability from Wearables: A Comparative Analysis Among Standard ECG, a Smart Shirt and a Wristband

Heart Rate Variability (HRV) derived from standard one-lead electrocardiography (ECG) was compared with HRV computed by a commercial ECG shirt and with the inter-beat-intervals (IBI) measured by a research-grade photoplethysmographic (PPG) wristband. Signals from 8 subjects were recorded in three experimental phases: during sit; in upright position ("stand"); during controlled respiration. HRV and IBI from both the wearables were computed online (i.e. during the experiment) and stored for subsequent analyses, while the standard ECG was processed offline with state-of-the-art methods to obtain a clean reference HRV. Shirt and wristband signals accuracies were assessed, with respect to the reference HRV, through a between-phase and a beat-to-beat analyses. The former considered several time and frequency domain parameters; the latter was carried out through the Bland-Altman method. Time and frequency domain parameters computed from shirt HRV resulted very similar to the ones extracted from the reference HRV and generally more accurate than the parameters computed from wristband IBI. The Bland-Altman analysis showed that wristband IBI is significantly different from ECG-derived HRV, especially during "stand". Therefore, our results support the idea that some care should be paid in the interpretation of online PPG-derived IBI, while HRV measures online-derived from ECG-shirts seem to be more reliable in the analyzed conditions. The high number of missing beats also suggest that the design of wristband devices should be taken into account to reduce the rate of incorrect measurements, by maximizing sensor adhesion to the skin.