Diagnosis of Atrial Fibrillation Using Machine Learning With Wearable Devices After Cardiac Surgery: Algorithm...

Area of Science:

• Cardiology
• Biomedical Engineering
• Medical Devices

Background:

• Wearable devices using photoelectric pulse wave technology are being explored for atrial fibrillation (AF) detection in clinical settings.
• Previous research has attempted AF detection with wearable sensors, highlighting the need for robust diagnostic methods.

Purpose of the Study:

• To develop an algorithm for immediate detection of paroxysmal AF using an Apple Watch with photoplethysmography (PPG).
• To apply machine learning to PPG pulse data from patients undergoing cardiac surgery for AF diagnosis.

Main Methods:

• 80 patients undergoing cardiac surgery were monitored postoperatively using telemetry-monitored ECG and an Apple Watch.
• A diagnostic algorithm was developed using machine learning on pulse rate data from the Apple Watch's PPG sensor.
• AF diagnosis was confirmed by qualified physicians using ECG data.

Main Results:

• Of 79 analyzed patients, 27 (34.2%) developed AF, with 199 AF events observed.
• Pulse rate data from the Apple Watch showed strong to very strong correlation with ECG-confirmed AF events.
• The machine learning algorithm achieved a diagnostic accuracy of 0.9416 (sensitivity 0.909, specificity 0.838).

Conclusions:

• The Apple Watch was safely worn by patients post-cardiac surgery for pulse rate monitoring.
• Despite some discrepancies between PPG and ECG heart rate measurements, the study demonstrates the potential for early AF detection using PPG data.
• This approach shows clinical applicability for the early detection of AF using wearable PPG sensors.