Abstract
Arterial stiffness is a key predictor of cardiovascular mortality. This study utilizes ultrasound-based Pulse Wave Imaging (PWI) and Vector Flow Imaging (VFI) to track vessel wall displacement caused by arterial pulse wave propagation and blood flow velocity at a high frame rate (3.3 kHz) to estimate localized arterial wall stiffness through an Inverse problem setting. We propose a physics-informed neural network (PINN) model to assess spatially varying arterial stiffness, integrating linearized 1D differential equations of pulse wave propagation in flexible tubes. Its effectiveness is validated through in silico data from a finite-difference 1D simulation and a plaque phantom. The proposed PINN model accurately captures localized compliance variations, reflecting arterial wall stiffness in both in-silico and phantom experiments. Future research will aim to incorporate non-linearities in the governing equations and expand the neural network to accommodate higher-dimensional spatial and temporal data.