Feature and voxel fidelity constraints improve the accuracy of direct inversion quantitative susceptibility mapping in deep gray matter

Abstract

We have designed a novel QSM algorithm that addresses some of the limitations of existing techniques that combine the background removal and dipole inversion steps in a single step. We propose that the solution to the direct inversion problem can be aided by an iterative k-space algorithm and the inclusion of a priori information that represents feature-based and voxel-fidelity-based constraints. The considered approach, when compared with other techniques, resulted in a more accurate depiction of the susceptibility in high susceptibility deep gray matter (dGM) structures without sacrificing performance in regions like the cortex of the brain.