Reconstructing magnetic resonance spectroscopic image using spatial domain prior

Magnetic resonance spectroscopic (MRS) images are low in resolution and often contain low frequency artifacts due to Fourier reconstruction of incomplete sets of data in k (frequency) space. The problem of improving MRS reconstructions can be approached in several ways, resulting in a number of apparently different formulations. These include the extrapolation of missing data samples, and also model-based methods that incorporate prior information from both the spatial and frequency domains. Model-based approaches can be shown to belong to a well-known class of image restoration problems. In the paper, the general MRS image reconstruction problem is reviewed, several approaches to the solution are given, and a particular method is presented that uses finite element models to embed spatial domain priors