Philip A Corrado1, Rafael Medero2, Kevin M. Johnson1,3, Christopher J François 3, Alejandro Roldán-Alzate2,4, and Oliver Wieben1,3
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, 2Mechanical Engineering, University of Wisconsin-Madison, Madison, WI, United States, 3Radiology, University of Wisconsin-Madison, Madison, WI, United States, 4Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, 2Mechanical Engineering, University of Wisconsin-Madison, Madison, WI, United States, 3Radiology, University of Wisconsin-Madison, Madison, WI, United States, 4Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States
We compared
3D-radial and stack-of-stars trajectories for 4D flow MRI in a left ventricle
model versus a reference dataset acquired with particle imaging velocimetry.
3D-radial matched the reference better, suggesting it is better for accelerated
left ventricular 4D flow MRI.
Figure 2: Velocity
vector comparison for PIV (left), 3D-radial MRI (middle), and SOS MRI (right)
images at two time points. Both MRI image sets were reconstructed with 5
minutes worth of data using wavelet-transform compressed sensing. The PIV
images look smoothest, but the patterns are similar in all images with some remaining
local differences.
Figure 3: Average
velocity errors in the LV comparing 4D flow MRI acquisitions to PIV (top row)
and to each trajectory’s most densely sampled acquisition (bottom row). Average
velocity error was computed by taking the average magnitude of the velocity
vector difference between test and reference image for all voxels in the LV. 2V
signifies dual-Venc and 1V signifies single-Venc. The Cartesian images were
included on both left (gridding) and right (CS Wavelet) graphs even though they
were reconstructed on-line with a proprietary kt-ARC reconstruction.