Michael Andrew Sharman1, Julien Cohen-Adad2,
Maxime Descoteaux3, Arnaud Mess4,5, Habib Benali4,5,
Stphane Lehericy6,7
1UMR-S975, CRICM-UPMC/Inserm, Paris,
le-de-France, France; 2Athinoula A. Martinos Center for
Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School,
United States; 3Department of Computer Science, Sherbrooke
University, Qubec, Canada; 4UMR-S678, UPMC/Inserm, Paris, France;
5IFR49, Paris, France; 6Centre for NeuroImaging
Research (CENIR), Hospital Piti-Salptrire , Paris, France; 7UMR-S975,
CRICM-UPMC/Inserm, Paris, France
Corrupted
images within acquired diffusion weighted MRI data can have an impact on the
estimation of the tensor (in diffusion tensor imaging) and diffusion ODF (in
q-ball imaging). In this study we performed a series of simulations and real
data analyses to quantify this impact on derived metrics such as fractional
anisotropy (FA) and generalised FA. From the results of these invetigations,
we propose processing strategies to detect and correct corruption artifacts
arising from large, unpredicatable signal variations.