Matching Pursuit Iterative Dipole Based Filter of Background Fields in Phase Imaging

Figure 3 three orthogonal slices of partial brain field maps (a) before and (b) after filtering. For correct background field removal it was necessary to pad the FOV by a factor of 2.5, 2.5 and 6 in the x, y and z directions respectively. The calculation of the background was performed on a matrix of resolution 3x3x1.5mm. The parameters of this specific filtering were t=0.3, n=20 Figure 2 Field in a transverse slice generated in the brain by magnetic susceptibility: (a) outside the brain, out; (b) inside the brain, brain; (c) and their combination. (d)Plot systematizing the performance of the different background field filtering algorithms on the data shown on the left. An ideal filter would be positioned on the top left of the plot. The different curves on the top left of the graph correspond to different parameter choices for the Matching Pursuit algorithm: using different values of n; different down sampling, Various points in the high pass filter curve, polynomial fit and conjugate gradient method correspond to the width of the Gaussian filter (1-10 pixels), order of the polynomial (0-4) and level of regulatization of the CG method (1-1e10) 4313 Matching Pursuit Iterative Dipole Based Filter of Background Fields in Phase Imaging