Assessment of a standalone photoplethysmography (PPG) algorithm for detection of atrial fibrillation on wristband-derived data. (December 2020)

• Record Type: Journal Article
• Authors: Selder, JL; Proesmans, T; Breukel, L; Dur, O; Gielen, W; van Rossum, AC; Allaart, CP
• Is Part Of: Computer methods and programs in biomedicine. Volume 197(2020)
• Publication Date: 2020-12
• Subjects: Smartwatch -- Wristband -- Atrial fibrillation -- Photoplethysmography -- One-lead ECG -- Algorithm -- Screening -- Monitoring -- Long-term -- Motion artefacts -- Quality assessment; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2020.105753
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: The fibricheck AF algorithm can accurately detect AF from wristband-derived PPG data in an elderly population, thus showing feasibility of non-invasive long-term AF screening and monitoring. Accuracy of AF detection by this PPG algorithm/wristband combination is comparable with 1-lead ECG devices. To prevent false positives, PPG algorithms need to be very strict in quality assessment. In this study 24% of the recordings are marked as insufficient quality. Abstract: Introduction: Atrial fibrillation (AF) is the most common cardiac arrhythmia in the developed world. Using photoplethysmography (PPG) and software algorithms, AF can be detected with high accuracy using smartphone camera-derived data. However, reports of diagnostic accuracy of standalone algorithms using wristband-derived PPG data are sparse, while this provides a means to perform long-term AF screening and monitoring. This study evaluated the diagnostic accuracy of a well-known standalone algorithm using wristband-derived PPG data. Materials and Methods: Subjects recruited from a community senior care organization were instructed to wear the Wavelet PPG wristband on one arm and the Alivecor KardiaBand one-lead-ECG wristband on the other. Three consecutive measurements (duration per measurement: 60 s for PPG and 30 s for one-lead ECG) were performed with both devices, simultaneously. The PPG data were analyzed by the Fibricheck standalone algorithm and the ECG data by the Kardia algorithm. The results were compared to a reference standard (interpretation of the one-lead ECG by two independent cardiologists). Results: A total of 180 PPGs and one-lead ECGs were recorded in 60 subjects, with a mean age of 70±17. AF was identified in 6 (10%) of the users, two users (3%) were not classifiable by the PPG algorithm and 1 user (2%) was not classifiable by the one-lead ECG algorithm. The diagnostic performance (sensitivity/specificity/positive predictive value/negative predictive value/accuracy) on user level was 100/96/75/100/97% for the PPG wristband and 100/98/86/100/98% for the one-lead ECG wristband. Conclusions: In a small real-world cohort of elderly people, the standalone Fibricheck AF algorithm can accurately detect AF using Wavelet wristband-derived PPG data. Results are comparable to the Alivecor Kardia one-lead ECG device, with an acceptable unclassifiable/bad quality rate. This opens the door for long-term AF screening and monitoring.