Vascular Aging Estimation Based on Artificial Neural Network Using Photoplethysmogram Waveform Decomposition:...

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Artificial Intelligence in Medicine

Background:

• Arterial stiffness assessment is crucial for evaluating arterial aging.
• Existing photoplethysmogram (PPG) methods lack reliability due to waveform feature limitations and regression model expressivity.
• A robust, noninvasive PPG-based arterial stiffness assessment is needed.

Purpose of the Study:

• Identify novel PPG features from incident and reflected waves correlated with vascular aging.
• Develop an artificial neural network (ANN) regression model for vascular age estimation using these features.

Main Methods:

• Acquired PPG waveforms from 757 participants.
• Decomposed PPG waveforms into incident and reflected waves using Gaussian mixture models.
• Defined 78 morphological features from decomposed waves and developed an ANN regression model.

Main Results:

• Reflected wave amplitude and PPG skewness showed strong correlation with chronological age.
• The ANN model estimated age with a root mean square error of 10.0 years.
• Estimated age and real age demonstrated a significant correlation (r=0.63, P<.001).

Conclusions:

• Features derived from PPG reflected waves and skewness are valuable for vascular aging assessment.
• The developed ANN model shows promise for noninvasive vascular age estimation.