Purpose: 4D-flow MRI obtains a time-dependent 3D velocity field; however, its use for the calculation of higher-order parameters is limited by noise. We present an algorithm for denoising 4D-flow data.
Theory and methods: By integrating a velocity field and eliminating streamlines in noisy flow, depicted by high curvature, a denoised dataset may be extracted. This method, defined as the velocity field improvement (VFIT) algorithm, was validated in an analytical dataset and using in vivo data in comparison with a computation fluid dynamics (CFD) simulation. As a proof of principal, wall shear stress (WSS) measurements in the descending aorta were compared with those defined by CFD.
Results: The VFIT algorithm achieved a >100% noise reduction of a corrupted analytical dataset. In addition, 4D-flow data were cleaned to show improved spatial resolution and near wall velocity representation. WSS measures compared well with CFD data and bulk flow dynamics were retained (<2% difference in flow measurements).
Conclusion: This study presents a method for denoising 4D-flow datasets with improved spatial resolution. Bulk flow dynamics are accurately conserved while velocity and velocity gradient fields are improved; this is important in the calculation of higher-order parameters such as WSS, which are shown to be more comparable to CFD measures. Magn Reson Med 78:1959-1968, 2017. © 2016 International Society for Magnetic Resonance in Medicine.
Keywords: 4D-flow; denoising; wall shear stress.
© 2016 International Society for Magnetic Resonance in Medicine.