A curated arterial stiffness dataset for vascular age prediction in China.

Abstract

Arterial stiffness is an important biomarker of cardiovascular health, and vascular age (VA) prediction provides additional value beyond chronological age. Here we present a curated arterial stiffness dataset comprising 36,223 participants aged 30-80 years from China. To benchmark its utility for VA modelling, we evaluated the Klemera-Doubal Method (KDM) and six Artificial Intelligence (AI) models: multiple linear regression, LASSO, random forest, support vector regression, XGBoost, and a deep neural network. Results showed that the dataset enables VA prediction using both statistical and learning-based approaches. Across both male and female cohorts, KDM showed the lowest prediction error under the current benchmark setting, while several nonlinear learning-based models achieved better performance than the linear baselines. Among the learning-based methods evaluated here, SVR and XGBoost showed comparatively strong performance. This dataset provides a useful open resource for vascular aging research, cardiovascular risk assessment, and methodological benchmarking.