(p. 485) Strategies for Patient–Control Comparison of Diffusion MR Data 

This chapter summarizes the main strategies for patient-control comparison of diffusion MRI data, including region of interest (ROI), histogram, voxel based analyses, and tractography. For each approach, a brief overview of the methods is given, followed by a discussion of the main limitations and advantages. In particular, it shows that the best strategy to extract quantitative information from diffusion data depends on the specific application. For example, ROI analysis is sensitive to small changes, particularly if concentrated in a small area of the brain; however, it is time-consuming and poorly reproducible, it requires an anatomical reference sharing the same geometry as the diffusion data, and the definition of clear guidelines; it also requires a strong hypothesis about the location of pathology. Histogram analysis is indicated when dealing with a diffuse disease, as it provides an assessment of the whole-brain without information on the location of pathology, and it requires an accurate procedure for CSF removal. An approach that conjugates the spatial specificity of ROI analysis with the possibility of assessing the whole brain is voxel-based analysis. A strong appeal of VB methods is the fact that while it might require long computational time, it requires minimal intervention from the user. Its reproducibility, however, strongly depends on the setting of the normalization and smoothing parameters. A diffusion tensor specific approach, named tract-based spatial statistics (TBSS), was recently developed. Its pros and cons are discussed. Finally, the most typical applications of diffusion tractography in clinical research are reviewed: tractography-based ROI definition, anatomical connectivity mapping, tract-shape definition and comparison, and connectivity-based parcellation.