Premium
User‐independent diffusion tensor imaging analysis pipelines in a rat model presenting ventriculomegalia: A comparison study
Author(s) -
Akakpo Luis,
Pierre Wyston C.,
Jin Chen,
Londono Irène,
Pouliot Philippe,
Lodygensky Gregory A.
Publication year - 2017
Publication title -
nmr in biomedicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.278
H-Index - 114
eISSN - 1099-1492
pISSN - 0952-3480
DOI - 10.1002/nbm.3793
Subject(s) - diffusion mri , computer science , artificial intelligence , fractional anisotropy , skeletonization , spatial normalization , white matter , interpretability , voxel , pattern recognition (psychology) , magnetic resonance imaging , medicine , radiology
Automated analysis of diffusion tensor imaging (DTI) data is an appealing way to process large datasets in an unbiased manner. However, automation can sometimes be linked to a lack of interpretability. Two whole‐brain, automated and voxelwise methods exist: voxel‐based analysis (VBA) and tract‐based spatial statistics (TBSS). In VBA, the amount of smoothing has been shown to influence the results. TBSS is free of this step, but a projection procedure is introduced to correct for residual misalignments. This projection assigns the local highest fractional anisotropy (FA) value to the mean FA skeleton, which represents white matter tract centers. For both methods, the normalization procedure has a major impact. These issues are well documented in humans but, to our knowledge, not in rodents. In this study, we assessed the quality of three different registration algorithms (ANTs SyN, DTI‐TK and FNIRT) using study‐specific templates and their impact on automated analysis methods (VBA and TBSS) in a rat pup model of diffuse white matter injury presenting large unilateral deformations. VBA and TBSS results were stable and anatomically coherent across the three pipelines. For VBA, in regions around the large deformations, interpretability was limited because of the increased partial volume effect. With TBSS, two of the three pipelines found a significant decrease in axial diffusivity (AD) at the known injury site. These results demonstrate that automated voxelwise analyses can be used in an animal model with large deformations.