A Modified Damped Richardson-Lucy Algorithm to Improve the Estimation of Fiber Orientations in Spherical Deconvolution

Introduction:. Standard spherical deconvolution approaches assume that HARDI signal can be modeled as a convolution between a common fiber response and a fiber orientation distribution function [1][2]. Recent spherical deconvolution methods improved the quality of the results introducing the non negative constraint of the solution [3][4][5]. It can be verified, however, that instability problems, such as spurious fiber orientations, could be related not only to noise robustness but also to signal contributes, such as from isotropic tissues, that are not properly taken into account. To develop new and more advanced fiber tracking techniques it is essential that fiber orientations are meaningful not only in main white matter (WM) fascicles but also in regions where partial volume effect between WM and isotropic tissue occurs. In this work, we proposed a modified damped version of the Richardson Lucy (RL) spherical deconvolution approach to reduce instabilities and spurious components also in these regions.