High Resolution Whole Head 3D Susceptibility Mapping at 7T: A Comparison of Multi-Orientation and Single Orientation Methods

Introduction: Phase images generated using gradient echo techniques at high field strength show excellent contrast related to the variation of magnetic susceptibility across different brain tissues [1] (Fig. 1a&d). Extraction of accurate anatomical information from phase images is however made difficult by the non-local relationship between the field perturbation, which underlies the phase variation, and the associated susceptibility distribution. Several approaches for calculating magnetic susceptibility from magnetic field measurements made using MRI have been proposed based on the deconvolution of field maps [2-5]. However, this deconvolution is an ill-posed problem requiring careful conditioning, for which a number of methods have been developed, including: (i) combination of phase data acquired at multiple orientations [2, 3]; (ii) k-space thresholding of data acquired at a single orientation [3,4]; (iii) incorporation of structural information using corresponding modulus data (Fig.1b&e) acquired at a single orientation [5]. While the multi-orientation method gives excellent results the need for head rotation is a limitation, particularly in studies involving patients. In contrast, the single orientation methods are simple to implement, but can be susceptible to artefacts. Here, we investigate the efficacy of applying the different conditioning strategies to high resolution 3D data acquired at 7T, first to images of a specially created phantom, and then to in-vivo data spanning the whole head, and compare the results of each method in regions showing significant susceptibility variation.