Meeting Banner
Abstract #2493

Relevance of Morphological Binary Information for L2 and L1 Total Variation Methods in Quantitative Susceptibility Mapping and Reconstruction Quality Assessment Without Presence of the Ground Truth

Diana Khabipova1, Yves Wiaux2, 3, Rolf Gruetter4, 5, Jos P. Marques1, 5

1LIFMET, EPFL, Lausanne, Switzerland; 2Electrical Engineering, EPFL, Lausanne, Switzerland; 3Radiology, UniGe, Geneva, Switzerland; 4LIFMET, cole Polytechnique Fdrale de Lausanne, Lausanne, Switzerland; 5Radiology, UNIL, Lausanne, Switzerland

Phase imaging has been demonstrated to achieve a good contrast between and within brain tissues at 7T. However, phase imaging suffers from a non-local contrast variation which can be overcome by calculating the underlying magnetic susceptibility maps. As this problem is ill-posed, many regularization methods have been proposed over the past years. In this abstract we do a thorough comparison of some of these methods, focus on the impact of the prior information on the reconstructed susceptibility maps and propose a method to evaluate the quality of the susceptibility reconstruction in the absence of a ground truth.

Abstract PDF Poster PDF

Keywords

absence achieve acquisition added agreement alternative amplitudes approved assessment assumptions available avoiding axial axis background binary brain calculating channel close coil coils combined committee compartments compensated consistency constant consuming contains contaminated contrast control convolution correction cosmos dependence dependent deviation dipole distribution echoes edge electrical engineering error estimation ethics evaluate evaluated even excluding experimental factorization field fields flow focus foundation generated good gradient ground head healthy highest important included includes independent least local lowest magic magnitude makes many mapping maps marques mask matrix measured methodologies methodology minimize minimizes moreover morphological noise norm nova numerical observation optimum orientations oriented overcome past phantom posed power prior priors problem profit promising protocol quality quantitative radiology random receive reconstructed reconstruction reconstructions reduction regularization regularized remains removed represents rows scanned scanner selection seven sharp simulated simulations solution space square subject subscript substantially suffers supported susceptibility third thorough threshold throughout truth unwrapped usage variation volunteer white years zero