The importance of priors for l2 regularization and total variation methods in quantitative susceptibility mapping

Phase imaging has been demonstrated to offer a good contrast between and within brain tissues at 7T[1] with iron and myelin concentration being amongst the main modulators of the observed contrast due to their para- and dia-magnetic properties.Phase imaging suffers from a non-local contrast variation which can be overcome by calculating the underlying magnetic susceptibility maps[2].As this problem is ill-posed, many regularized methods have been proposed over the past years[2,3,4].The regularized single-orientation (RSO)[2] based on the l2 regularisation and the morphology enabled dipole inversion (MEDI)[3] method based on the l1 total variation incorporate prior knowledge of the expected edges taken from the magnitude image.In this abstract a systematic evaluation is performed of a l2 RSO method[2,5] and a l1 total variation method (TV)[3] using numerical data and different morphology priors.