Smartwatch Based Atrial Fibrillation Detection from Photoplethysmography Signals.

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Signal Processing

Background:

• Atrial fibrillation (AF) detection is crucial for preventing stroke and requires continuous, non-invasive monitoring.
• Photoplethysmogram (PPG) signals from wrist-worn devices offer a promising avenue for long-term AF monitoring.
• Existing methods often struggle with motion artifacts corrupting PPG signals.

Purpose of the Study:

• To develop a novel method for detecting AF from wristwatch PPG signals.
• To automatically distinguish between clean and corrupted PPG segments.
• To enable reliable, continuous, and non-invasive AF monitoring.

Main Methods:

• Utilized accelerometer data and time-frequency analysis (variable frequency complex demodulation) to detect motion and noise artifacts in PPG signals.
• Extracted features including root mean square of successive differences and sample entropy from beat-to-beat intervals of clean PPG signals.
• Employed a UMass dataset with 20 subjects for algorithm validation.

Main Results:

• Achieved high performance in AF detection: 96.15% sensitivity, 97.37% specificity, and 97.11% accuracy.
• Successfully distinguished between clean and corrupted PPG signal segments.
• Demonstrated the algorithm's efficacy on a real-world dataset.

Conclusions:

• The proposed method offers a practical and reliable approach for AF monitoring using wrist-worn PPG devices.
• Automatic artifact detection and removal enhance the accuracy of AF detection from PPG signals.
• This technology holds significant potential for widespread clinical application in cardiovascular health.