Post-Processing of R Wave Detection to Enhance Heart Rate Variability Analysis

Heart rate variability (HRV) is a key biomarker reflecting autonomic nervous system activity associated with various health conditions. To analyze HRV, it is necessary to detect R waves from an electrocardiogram (ECG) and calculate R-R intervals (RRIs). However, completely eliminating R wave misdetections is technically challenging, and misdetected R waves cannot be always identified solely based on the duration of the RRIs. To achieve high-precision HRV-based health monitoring using wearable devices in daily life environments, RRI editing is required to reject misdetected R waves and interpolate missing values. As its initial step, in this study, we aimed to identify misdetected R waves utilizing the information available after R wave detection: ECG amplitude and detected position. The evaluation results comparing nine conditions, including three different machine learning algorithms and three types of preprocessing, showed that non-linear algorithms combined with either min-max normalization or Z-score normalization achieved higher classification performance, with a Matthew's correlation coefficient of 0.86.Clinical Relevance-This study will contribute to the reliability of HRV analysis using wearable devices in daily life environments.