Deriving high-resolution velocity maps from low-resolution fourier velocity encoded MRI data

Abstract

Fourier velocity encoding (FVE) is a promising magnetic resonance imaging (MRI) method for measurement of cardiovascular blood flow. FVE provides considerably higher SNR than phase contrast imaging, and is robust to partial-volume effects. FVE data is usually acquired with low spatial resolution, due to scan-time restrictions associated with its higher dimensionality. Thus, FVE is capable of providing the velocity distribution associated with a large voxel, but does not directly provides a velocity map. Velocity maps, however, are useful for calculating the actual blod flow through a vessel, or for guiding computational fluid dynamics simulations. This work proposes a method to derive velocity maps with high spatial resolution from low-resolution FVE data using a hyper-Laplacian prior deconvolution algorithm. Experiments using numerical phantoms, as well simulated spiral FVE data derived from real phase contrast data, acquired using a pulsatile carotid flow phantom, show that it is possible to obtain reasonably accurate velocities maps from low-resolution FVE distributions.

Fourier Velocity Encoded MRI: Acceleration and Velocity Map Estimation

“…The velocity maps estimated from the simulated low spatial resolution FVE data are very similar (qualitatively) to the reference map. At first glance one can say that the velocity map obtained using the technique proposed by Rispoli and Carvalho [55] (Figure 7c) is more similar to the acquired PC-MRI velocity map. However the error images show that the velocity map obtained using the technique proposed in this work ( Figure 7b) was more accurate than the one obtained with the other method ( Figure 7c).…”

“…Finally, the proposed method measured SER, relative to the PC reference, was 44.63 dB while the technique proposed in Rispoli and Carvalho [55] achieved 28.68 dB. Showing that the proposed optimization given by Eq.…”

“…This is because FVE has considerably higher SNR than PC in general, due to its higher dimensionality and larger voxel size. Finally, two different spiral FVE distributions were obtained for each slice of the volume with Δr =2mmspatialresolution:one using the proposed method and the other one using the method proposed by Rispoli and Carvalho [55]. The velocity resolution was set to Δv ¼ 10 cm/s, over a 120 cm/s velocity field-of-view.…”

“…Validation experiment using a pulsatile carotid flow phantom: (a) reference phase contrast velocity map, measured at the phantom's bifurcation; (b) velocity map estimated from the simulated low-resolution spiral FVE data with Δr = 2 mm spatial resolution with the proposed method (and associated error percentages); and (c) velocity map estimated from the simulated low-resolution spiral FVE data with Δr = 2 mm spatial resolution with the method proposed by Rispoli and Carvalho[55] (and associated error percentages).…”