Continuous heart rate measurements in patients with cardiac disease: Device comparison and development of a novel artefact removal procedure.

INTRODUCTION

Heart rate (HR) monitors could objectively measure physical activity intensity in patients with cardiac disease. However, thorough validation of HR monitors in cardiac populations during daily life, compared to gold-standard Holter monitoring, remains limited. Photoplethysmography (PPG)-based HR data provides near-continuous data, spanning longer periods, but improved algorithms to filter unreliable data are needed.

METHODS

This observational, prospective pilot study compared the accuracy of two wearables for HR monitoring (electrocardiogram [ECG]-based Polar H10 chest strap and PPG-based Fitbit Inspire 2 wrist tracker) against Holter monitoring in 15 patients with atrial fibrillation (AF), heart failure (HF) and coronary artery disease referred for cardiac rehabilitation (CR). All devices were worn simultaneously for 24 h. We developed and assessed an artefact removal procedure (ARP) using logistic regression machine learning models to detect unreliable PPG data.

RESULTS

The ECG-based chest strap showed a strong correlation ( = 0.94) and clinically acceptable errors (mean absolute error, MAE = 3.4 bpm; mean absolute percentage error, MAPE = 4.9%). Photoplethysmography data exhibited weaker correlation ( = 0.69) and higher errors (MAE = 8.3 bpm, MAPE = 14.3%), with highest accuracies in CR and lowest in HF and especially AF. After implementing the ARP, PPG-based HR data improved to a correlation of 0.75, with MAE of 7.2 bpm and MAPE of 12.4%. The procedure removed nearly one-third of unreliable data, achieving an 81% accuracy.

CONCLUSIONS

While ECG-based monitors provide HR data with clinical acceptable accuracy, PPG-based monitors present accuracy challenges. Our machine learning procedure showed potential to filter unreliable PPG-based HR data, which could help measure physical activity intensity in cardiac disease continuously.