Ultrasound super-resolved hemodynamic estimation in microvessel using physics-informed neural networks and data assimilation

Background and Objective Ultrasound super-resolution imaging (SRI) enables the visualization of microvascular structure and velocity, but enhancing the spatial resolution of instantaneous velocity field and simultaneously capturing pressure field remains challenging. Methods This study proposes a method combining physics-informed neural networks (PINN) with data assimilation to assist microvascular two-dimensional (2D) super-resolution velocity and pressure reconstruction in SRI. Specifically, long-time velocity vector set acquired via SRI is decomposed into short-time subsets, with vectors in each subset stacked and treated as simultaneous to enhance spatial information. These are then matched and fused with the hemodynamic simulation based on the SRI-derived structure and flow information via data assimilation, generating a new velocity field that effectively filling gaps in sparse measurements. This velocity is used to optimize the PINN encoded with the 2D Navier-Stokes equations to reconstruct the super-resolution velocity field and infer reliable pressure field. Results In vitro experiments validated the method’s performance and investigated the influence of the data amplification factor on the reconstruction accuracy, with the spatial vectors number increased by 6.48 times. Meanwhile, the super-resolution hemodynamic parameter reconstructions of rat brain microvessels and liver tumor peritumoral vessels aligned with the velocity measured by conventional SRI (rat brain vessels: radial resolution of 0.46 μm and axial resolution of 5.9 μm, liver tumor vessels: radial resolution of 5.5 μm and axial resolution of 123 μm), and the relative errors are 1.85% and 4.89%, respectively. Conclusions The proposed method reconstructs super-resolution microvascular velocity and pressure from sparse, inhomogeneous 2D SRI velocity data, showing powerful potential for aiding clinical diagnosis of microvascular diseases. (ClinicalTrials.gov