Generation of a Synthetized Database Based on Real Data for Vascular Age Estimation Through a Machine Learning Approach

Several studies have established reference values for the relationship between “healthy aging” and the evolution of arterial biomechanical variables, such as blood pressure, intima-media thickness, and pulse wave velocity. This study investigates the feasibility of estimating vascular age through the generation of a synthetic database that incorporates these cardiovascular parameters for individuals of different ages. To achieve this, a synthetic database (n=60,000) was created from relevant publications that included real data exclusively from healthy patients. This database was then utilized to estimate individual vascular age using different supervised machine learning models, understanding vascular age as the chronological age of a normotensive, non-pathologic individual with specific arterial biomechanical variable values. The models demonstrated promising performance, as indicated by the coefficient of determination (0.77), root mean square error (8.24 years), and mean absolute error (6.43 years), when assessed for their ability to predict chronological age in healthy individuals, where chronological age and vascular age are assumed to be equivalent. These results support the development of more personalized and accurate preventive measures, enabling early and potentially modifiable interventions in vascular health.