AUTHORS: Kebiri H, Lajous H, Yasser Aleman-Gomez Y, Girard G, Canales Rodriguez E, Tourbier S, Pizzolato M, Ledoux JB, Fornari E, Jakab A, Bach Cuadra M

12th International Workshop, CDMRI 2021, 13006: 12-22, Strasbourg, France, October 2021


Diffusion Magnetic Resonance Imaging (dMRI) has become widely used to study in vivo white matter tissue properties noninvasively. However, fetal dMRI is greatly limited in Signal-to-Noise ratio and spatial resolution. Due to the uncontrollable fetal motion, echo planar imaging acquisitions often result in highly degraded images, hence the ability to depict precise diffusion MR properties remains unknown.
To the best of our knowledge, this is the first study to evaluate diffusion properties in a fetal customized crossing-fiber phantom. We assessed the effect of scanning settings on diffusion quantities in a phantom specifically designed to mimic typical values in the fetal brain. Orthogonal acquisitions based on clinical fetal brain schemes were preprocessed for denoising, bias field inhomogeneity and distortion correction. We estimated the fractional anisotropy (FA) and mean diffusivity (MD) from the diffusion tensor, and the fiber orientations from the fiber orientation distribution function. Quantitative evaluation was carried out on the number of diffusion gradient directions, different orthogonal acquisitions, and enhanced 4D volumes from scattered data interpolation of multiple series. We found out that while MD does not vary with the number of diffusion gradient directions nor the number of orthogonal series, FA is slightly more accurate with more directions. Additionally, errors in all scalar diffusion maps are reduced by using enhanced 4D volumes. Moreover, reduced fiber orientation estimation errors were obtained when used enhanced 4D volumes, but not with more diffusion gradient directions.
From these results, we conclude that using enhanced 4D volumes from multiple series should be preferred over using more diffusion gradient directions in clinical fetal dMRI.

Download PDF


Module: |