Cerebral white matter analysis using diffusion imaging

Abstract

In this thesis we address the whole-brain tractography segmentation problem. Diffusion magnetic resonance imaging can be used to create a representation of white matter tracts in the brain via a process called tractography. Whole brain tractography outputs thousands of trajectories that each approximate a white matter fiber pathway. Our method performs automatic organization, or segmention, of these trajectories into anatomical regions and gives automatic region correspondence across subjects. Our method enables both the automatic group comparison of white matter anatomy and of its regional diffusion properties, and the creation of consistent white matter visualizations across subjects. We learn a model of common white matter structures by analyzing many registered tractography datasets simultaneously. Each trajectory is represented as a point in a high-dimensional spectral embedding space, and common structures are found by clustering in this space. By annotating the clusters with anatomical labels, we create a model that we call a high-dimensional white matter atlas.

(cont.) Our atlas creation method discovers structures corresponding to expected white matter anatomy, such as the corpus callosum, uncinate fasciculus, cingulum bundles, arcuate fasciculus, etc. We show how to extend the spectral clustering solution, stored in the atlas, using the Nystrom method to perform automatic segmentation of tractography from novel subjects. This automatic tractography segmentation gives an automatic region correspondence across subjects when all subjects are labeled using the atlas. We show the resulting automatic region correspondences, demonstrate that our clustering method is reproducible, and show that the automatically segmented regions can be used for robust measurement of fractional anisotropy.