Comparison between Synb0-DisCo and Fieldmap-based methods to correct for distortion artifacts in diffusion MRI
Abstract
Diffusion-weighted images acquired with echo planar imaging (EPI) are sensitive to static magnetic field inhomogeneities that can cause distortion artifacts in the phase encoding direction. These artifacts are detrimental for the analyses and several methods have been developed to correct them. Some methods require the acquisition of a dedicated B 0 map like the fieldmap-based method and the state-of-theart "blip-up/blip-down" method. Others do not, like methods using non-rigid registration to a structural image, but they show weaker results. Recently, a method called Synthesized b0 Distortion Correction (Synb0-DisCo) that uses deep learning to perform corrections without additional data (Schilling et al. 2020) has been developed. This study evaluates the Synb0-DisCo algorithm against the other mentioned methods. We performed comparisons on the displacement fields, the $b$ = 0 (diffusion-unweighted) images, and between the $b$ = 0 images and the T1-weighted images; using monocentric data from 50 healthy subjects. We found that, in this context, the methods using fieldmaps outperform the Synb0-DisCo method in most respects.
Fichier principal
papier_synb0_preprint (1).pdf (1.84 Mo)
Télécharger le fichier
poster_isbi.pdf (1.34 Mo)
Télécharger le fichier
Origin | Files produced by the author(s) |
---|