MODEL-BASED MR IMAGE RECONSTRUCTION WITH COMPENSATION FOR THROUGH-PLANE FIELD INHOMOGENEITY

To improve image quality in susceptibility-weighted MR imaging, it is important to correct for the effects of field inhomogeneity. In particular, susceptibility differences between air and tissue induce magnetic field nonuniformity; often those susceptibility effects have nonzero through-plane gradients that lead to spin dephasing across the slice within each voxel. If uncorrected, these through-plane gradients cause signal loss in the reconstructed images. Several methods exist for reconstructing MR images with compensation for field inhomogeneity, but most of these methods, even the model-based iterative ones, treat the inhomogeneity within each voxel as being a constant, thus ignoring the through-plane gradient effects. This paper describes a model-based iterative method for reconstructing MR images with compensation for field inhomogeneity that accounts for the slice profile and the through-plane gradients of the field inhomogeneity (assumed to be determined by a pre-scan). In particular, this paper describes an accelerated algorithm for implementing the forward model and its adjoint as needed in a conjugate gradient algorithm for iterative MR image reconstruction.