Super-Resolving 4D Flow MR Images for Accurate Hemodynamic Calculations: In- Vitro Validation and Clinical Application

Abstract

In this study, we present two in-vitro measurements of a carotid artery phantom model using four-dimensional magnetic resonance flow imaging (4DMR) in a 3 Tesla (3T) and a 7T scanner, respectively, and show that the accuracy of velocity measurements can be enhanced by the previously proposed super-resolution (SR) model 4DFlowNet. To demonstrate the clinical relevance, we trained a segmentation model based on nnU-Net using a clinical dataset and applied the same data pipeline on the luminal flow of the segmented carotid arteries. By comparing the calculated wall shear stresses before and after the SR process, we conclude that the SR model can be a competent tool to improve the diagnostic power of 3T MR images.