Validation of a Deep Learning Algorithm for Continuous, Real-Time Detection of Atrial Fibrillation Using a Wrist-Worn...

Area of Science:

• Cardiology
• Biomedical Engineering
• Digital Health

Background:

• Consumer wearable devices offer periodic atrial fibrillation (AF) detection but lack clinical-grade, continuous monitoring capabilities.
• Current devices are cleared for prediagnostic use, not clinical decision-making, highlighting an unmet need for medical-grade solutions.
• Long-term, continuous AF monitoring is crucial for diagnosis and management.

Purpose of the Study:

• To evaluate the performance of a wrist-worn device with ECG and photoplethysmography (PPG) for AF detection and burden estimation.
• To assess the accuracy of a PPG-based convolutional neural network (CNN) algorithm for AF monitoring.
• To compare the device's performance against a reference ECG monitor.

Main Methods:

• A prospective multicenter study enrolled 117 patients with paroxysmal AF.
• A wrist-worn device (Verily Study Watch) with lead-I ECG and PPG was used for continuous monitoring.
• A 14-day continuous ECG monitor (Zio XT) served as the reference standard for algorithm evaluation.

Main Results:

• The wrist-worn device achieved 96.1% sensitivity and 98.1% specificity for AF detection in 15-minute intervals.
• PPG-derived AF burden estimation showed high correlation (R²=0.986) with the reference device.
• Median watch wear time was 18.3 hours/day across 111 subjects with evaluable data.

Conclusions:

• A PPG-based CNN in a wrist-worn device provides clinical-grade AF detection and burden estimation.
• Continuous monitoring with this wearable technology supports AF diagnosis and clinical management.
• This advancement addresses the need for reliable, long-term AF monitoring solutions.