Pulse Wave Velocity Estimation Using Photoplethysmogram-Based Limited Penetrable Weighted Visibility Graph Features.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Pulse Wave Velocity (PWV) is a crucial indicator of arterial stiffness and cardiovascular health.
• Accurate PWV estimation is vital for non-invasive cardiovascular risk assessment.
• Traditional methods for PWV estimation can be complex or require specialized equipment.

Purpose of the Study:

• To propose a novel data-driven model for Pulse Wave Velocity (PWV) estimation.
• To leverage Limited Penetrable Weighted Visibility Graphs (LPWVG) derived from photoplethysmogram (PPG) waveforms.
• To evaluate the effectiveness of machine learning models using extracted PPG features for PWV estimation.

Main Methods:

• Construction of four distinct LPWVGs using diverse weighted methods from PPG waveforms.
• Extraction of various features from PPG signals, including 2D Semi-classical Signal Analysis (SCSA)-based, frequency-based, and shape-based features.
• Inputting extracted features into different machine learning models for PWV estimation.

Main Results:

• The proposed LPWVGs and feature extraction methods demonstrated effectiveness in PWV estimation.
• Performance evaluation using both in-silico and real PPG pulse wave data confirmed the model's accuracy.
• The study provides strong evidence for the feasibility of the proposed data-driven approach.

Conclusions:

• The developed data-driven model utilizing LPWVGs from PPG signals is effective for accurate PWV estimation.
• This method shows significant potential for advancing non-invasive cardiovascular diagnostics.
• The approach offers a promising tool for biomedical applications requiring precise PWV measurement.