Vascular Ageing Prediction with Visibility Graphs

Vascular ageing is a crucial metric of cardiovascular system health. Numerous studies have explored predicting vascular ageing using Photoplethysmography (PPG) signals and employing deep learning techniques. Nevertheless, these studies exhibit limitations such as reliance on human-involved processing, susceptibility to signal corruption, and dependence on signal amplitude. In this study, we propose a novel approach for detecting vascular ageing with PPG signals. The proposed algorithm combines visibility graphs with deep learning, offering a robust estimation with affine-invariant and amplitude-independent characteristics. We tested our method on multiclass classification, binary classification, and age regression. Our model has demonstrated superior performance compared to other well-established baselines.