HeartView: An Extensible, Open-Source, Web-Based Signal Quality Assessment Pipeline for Ambulatory Cardiovascular Data

Wearable sensing systems enable peripheral physiological data to be collected repeatedly in naturalistic settings. However, the ambulatory nature of wearable biosensors predisposes them to common signal artifacts that researchers must address before analysis. Signal quality assessment procedures are time-consuming and non-standardized across research teams, and transparent reporting of custom, closed-source pipelines needs improvement. This paper presents HeartView, an extensible, open-source, web-based signal quality assessment pipeline that visualizes and quantifies missing beats and invalid segments in heart rate variability (HRV) data obtained from ambulatory electrocardiograph (ECG) and photoplethysmograph (PPG) signals. We demonstrate the utility of our pipeline on two datasets: (1) 34 ECGs recorded with the Actiwave Cardio from children with and without autism, and (2) 15 sets of ECGs and PPGs recorded with the RespiBAN and Empatica E4, respectively, from healthy adults in the publicly available WESAD dataset. Our pipeline demonstrates interpretable group differences in physiological signal quality. ECGs of children with autism contain more missing beats and invalid segments than those without autism. Similarly, PPG data contains more missing beats and invalid segments than ECG data. HeartView has a graphical user interface in the form of a web-based dashboard at https://github.com/cbslneu/heartview.