Development of Vascular Aging Assessment Model Based on Photoplethysmogram Incident and Reflected Wave Characteristics

This study aims to develop a regression model for assessment of vascular aging based on characteristics of photoplethysmogram incident and reflected wave decomposed by Gaussian mixture model. We obtained photoplethysmogram from 757 participants and defined 16 primary and 20 combined features that were created from the morphological characteristics of incident and reflected wave in temporal and spatial domain. In correlation analysis, the features from amplitude of reflected wave and skewness of PPG had the highest correlation coefficient with subject’s real age. The vascular aging assessment models were developed by linear and non-linear regression. As a result, the 2SUPnd/SUP order polynomial model developed with 5 features based on skewness of PPG, kurtosis of incident wave, time of incident wave’s peak, area of incident wave and amplitude of reflected wave’s peak showed 9.6 years of root mean square error, outperforming the linear model.