Complex Domain Inverse Problem Formulation for 4D Flow MRI Super-Resolution

Abstract

Four-dimensional flow magnetic resonance imaging measures blood velocity fields in three dimensions throughout the cardiac cycle and is increasingly used to assess hemodynamics. However, these measurements are limited by low signal-to-noise ratios and limited spatial resolution, which can compromise their accuracy and reliability. To address these limitations, we investigate the feasibility of spatial resolution enhancement through a complex-domain inverse problem formulation inspired by MRI acquisition principles. Three inversion methods are evaluated: a closed-form approach (FSR), a Plug-and-Play alternating direction method of multipliers (PnP-ADMM), and a pretrained network that takes the forward operator as input (RAM). On simulated data, all three methods outperform 4DFlowNet at x2 and x4 super-resolution factors. On aortic phantom data at x2 super-resolution, PnP-ADMM and RAM achieve lower root-mean-square error than 4DFlowNet in almost all cases. These results highlight the potential of the proposed formulation and support further development and validation.