Magnetic resonance imaging maps of velocity were acquired with a 1.5-T system in 10 subjects in a plane perpendicular to the main pulmonary artery. Velocity images were successively acquired with a method developed from Fourier-encoding velocity imaging (FEVI) principles with eight gradient steps and one excitation, and with two-point phase-subtraction mapping. Reconstruction in FEVI was implemented by zero-filling interpolation around the eight gradient steps and then around the four central steps. The methods were compared by using estimates of noise in velocity measurements based on the difference between the experimental map and a smooth fitted map. For the same acquisition time, FEVI with four encoding steps was more precise in velocity measurements than phase mapping. Precision was further increased by the use of eight encoding steps, but acquisition time was doubled.

The final result of Fourier velocity mapping is a set of images, each representing the spatial distribution of spins at a given velocity. To acquire data in a short time, the number of encoding gradient steps must be as small as possible, but this can mean sacrificing velocity resolution. We used interpolation methods to obtain high velocity resolution with a small number of encoding steps involving linear interpolation from 16 encoding steps or more and zero-filling interpolation from two to eight encoding steps. Velocity measured by interpolated Fourier-flow encoding agreed well with values obtained using a calibrated phantom. A simulation of noise on the images of the phantom showed that, for a given acquisition time, increasing number of encoding steps in the Fourier flow encoding gave better precision for velocity measurement than did averaging identical signals in phase-mapping methods.