PPG-basis: A python-based toolbox for fast photoplethysmogram decomposition and reconstruction.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Physiological Monitoring

Background:

• Photoplethysmogram (PPG) data are crucial for extracting biometrics like heart rate and blood oxygenation.
• PPG waveform morphology contains valuable information for metrics such as vascular stiffness.
• Accurate reconstruction of noisy PPG data is essential for developing robust, noise-resistant models.

Purpose of the Study:

• To develop a unified toolbox for testing PPG data decomposition and reconstruction methods.
• To enable reliable extraction of complex biometrics from PPG morphology.
• To improve the accuracy and efficiency of PPG data analysis.

Main Methods:

• Reformulated PPG generation in the phase domain instead of the time domain.
• Utilized a look-up table to optimize computational processes.
• Developed a two-stage decomposition pipeline with three distinct kernels for PPG data reconstruction.

Main Results:

• Achieved an order-of-magnitude decrease in computational time compared to existing methods while maintaining high accuracy.
• Successfully decomposed and reconstructed experimental PPG data with injected noise.
• Extracted complex biometrics from N=50 subjects in the MIMIC-III clinical database.

Conclusions:

• The ppg-basis method accurately recovers morphological features with lower error than bandpass filtering on clean data.
• The framework maintains stable feature detection under noise injection, unlike filter-based methods.
• This approach can be used for developing noise-resistant PPG analysis models and testing their robustness.