User-independent diffusion tensor imaging analysis pipelines in a rat model presenting ventriculomegalia: A...

Area of Science:

• Neuroimaging
• Animal Models
• White Matter Injury

Background:

• Automated analysis of diffusion tensor imaging (DTI) offers unbiased processing of large datasets.
• Voxel-based analysis (VBA) and tract-based spatial statistics (TBSS) are whole-brain automated methods.
• Interpretability challenges exist with automated methods, particularly concerning smoothing and normalization, with limited data in rodent models.

Purpose of the Study:

• To assess the quality of three registration algorithms (ANTs SyN, DTI-TK, FNIRT) for automated DTI analysis in a rat pup model.
• To evaluate the impact of these registration algorithms on VBA and TBSS methods.
• To determine the utility of automated voxelwise analyses in animal models with significant deformations.

Main Methods:

• Utilized a rat pup model of diffuse white matter injury with large unilateral deformations.
• Applied three registration algorithms (ANTs SyN, DTI-TK, FNIRT) with study-specific templates.
• Performed automated whole-brain voxelwise analyses using VBA and TBSS.

Main Results:

• VBA and TBSS results demonstrated stability and anatomical coherence across the three registration pipelines.
• VBA interpretability was limited in areas of large deformation due to partial volume effects.
• Two of three pipelines identified a significant decrease in axial diffusivity (AD) at the known injury site using TBSS.

Conclusions:

• Automated voxelwise DTI analysis methods (VBA and TBSS) are viable in rodent models, even with substantial deformations.
• Registration algorithm choice impacts interpretability and sensitivity, particularly for VBA in highly deformed regions.
• TBSS showed promise in detecting injury-specific changes in axial diffusivity within this animal model.