Jessica Dubois1,2, Ghislaine
Dehaene-Lambertz1,3, Lucie Hertz-Pannier2,4, Giovanna
Santoro1, Jean-Franois Mangin, 3,5, Cyril Poupon,
3,5
1U562, Inserm, Gif-sur-Yvette, France; 2LBIOM,
CEA, Gif-sur-Yvette, France; 3IFR49, Paris, France; 4U663,
Inserm, Paris, France; 5LNAO, CEA, Gif-sur-Yvette, France
Diffusion
Tensor Imaging (DTI) offers the possibility to study the developing white
matter non-invasively. However, diffusion-weighted images obtained in non-sedated
infants are often corrupted with motion artifacts. We propose a
post-processing methodology which takes advantage of the high diffusion
orientation count and corrects these images before the computation of
diffusion maps. The strategy relies on three successive steps: two steps of
correction of corrupted slices (using decomposition on a spherical harmonics
basis), separated by a step of 3D motion registration. This approach was
validated on DTI data from 15 infants, by reliably evaluating the corpus
callosum maturation with tractography-based quantification.
Keywords