Impact of regularization factor on quantification of flow and turbulence in highly undersampled 4D flow MRI

We investigate the impact of regularization on the quantification of flow and turbulence in highly undersampled data. To overcome the uncertainties due to the lack of ground truth, patient-specific aortic geometry and inflow conditions were extracted from in-vivo 2D cine and 2D phase-contrast MRI and used to generate synthetic personalized ground truth flow fields. Simulation results were embedded into the corresponding patient-specific 4D flow MRI effectively resulting in personalized synthetic datasets with known aortic ground truth and realistic background. The reconstruction of multiple undersampled datasets showed two distinct optimal regularization ranges for the quantification of flow velocities and turbulence.