Fast Iterative Approximate Pseudo-Inverse Image Reconstruction from Data Acquired on Arbitrary k-Space Trajectories
Greengard L, Inati S
New York University
We present an efficient least squares method for the reconstruction of magnetic resonance images from data sampled non-uniformly in k-space. The procedure only requires knowledge of the sampling locations and can be applied to arbitrary trajectories in 2 or 3 dimensions, including those that are self-intersecting. The method is based on two fast algorithms, the non-uniform fast Fourier transform (NUFFT) and the fast sinc transform, combined with analytic weights, and an iterative procedure.