Fast, Iterative, Field-Corrected Image Reconstruction for MRI

Abstract

Magnetic field inhomogeneities cause distortions in the reconstructed images for non-cartesian k-space MRI (using spirals, for example). Several noniterative methods are currently used to compensate for the off-resonance during the reconstruction, but these methods rely on the assumption of a smoothly varying field map. Recently, iterative methods have been proposed that do not rely on this assumption and have the potential to estimate undistorted field maps, but suffer from prohibitively long computation times. In this abstract we present a min-max derived, time-segmented approximation to the signal equation for MRI that, when combined with the nonuniform fast Fourier transform, provides a fast, accurate field-corrected image reconstruction.