Detecting beats in the photoplethysmogram: benchmarking open-source algorithms.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiovascular Monitoring

Background:

• Photoplethysmogram (PPG) signals are crucial for pulse oximetry and wearables.
• Accurate heartbeat detection is fundamental for PPG signal analysis.
• Existing PPG beat detection algorithms lack comprehensive performance comparison.

Purpose of the Study:

• To establish a framework for designing and evaluating PPG beat detection algorithms.
• To assess algorithm performance across diverse use cases and physiological conditions.
• To identify the optimal PPG beat detection methods for clinical and wearable applications.

Main Methods:

• Developed a standardized framework for PPG beat detector assessment.
• Evaluated fifteen distinct beat detection algorithms.
• Utilized eight diverse datasets and electrocardiogram (ECG) for ground truth comparison.

Main Results:

• Eight algorithms achieved ≥90% F1 scores in resting conditions (hospital and wearable data).
• Performance decreased during exercise (55%-91% F1), in neonates (84%-96% F1), and with atrial fibrillation (AF) (92%-97% F1).
• MSPTD and qppg demonstrated superior and complementary performance.

Conclusions:

• MSPTD and qppg are recommended PPG beat detection algorithms.
• The developed framework and freely available resources facilitate algorithm selection.
• Performance is significantly influenced by patient demographics and physiological state.