Accuracy of four-dimensional phase-contrast velocity mapping for blood flow visualizations: a phantom study

Background: Time-resolved three-dimensional, three-directional phase-contrast magnetic resonance velocity mapping (4D PC-MRI) is a powerful technique to depict dynamic blood flow patterns in the human body. However, the impact of phase background effects on flow visualizations has not been thoroughly studied previously, and it has not yet been experimentally demonstrated to what degree phase offsets affect flow visualizations and create errors such as inaccurate particle traces. Purpose: To quantify background phase offsets and their subsequent impact on particle trace visualizations in a 4D PC-MRI sequence. Additionally, we sought to investigate to what degree visualization errors are reduced by background phase correction. Material and Methods: A rotating phantom with a known velocity field was used to quantify background phase of 4D PC-MRI sequences accelerated with SENSE as well as different k-t BLAST speed-up factors. The deviation in end positions between particle traces in the measured velocity fields were compared before and after the application of two different phase correction methods. Results: Phantom measurements revealed background velocity offsets up to 7 cm/s (7% of velocity encoding sensitivity) in the central slice, increasing with distance from the center. Background offsets remained constant with increasing k-t BLAST speed-up factors. End deviations of up to 5.3 mm (1.8 voxels) in the direction perpendicular to the rotating disc were found between particle traces and the seeding plane of the traces. Phase correction by subtraction of the data from the stationary phantom reduced the average deviation by up to 56%, while correcting the data-set with a first-order polynomial fit to stationary regions decreased average deviation up to 78%. Conclusion: Pathline visualizations can be significantly affected by background phase errors, highlighting the importance of dedicated and robust phase correction methods. Our results show that pathline deviation can be substantial if adequate phase background errors are not minimized.