Thesis Defense


"In-vivo Microstructural and Anatomical Analysis of Brain White Matter from Diffusion MRI"

Wenjin Zhou

Monday, July 23, 2012 at 11:00 A.M.

Lubrano Conference Room (CIT 4th floor)

I present new computational approaches toward the virtual histology of white-matter microstructures and new visualization and interaction techniques for identifying and segmenting white-matter anatomy. Together, these new techniques, which are based on diffusion magnetic resonance imaging (diffusion MRI), enhance the microstructural and anatomical analysis of brain white matter to localize neurological changes resulting from disease and degeneration in two ways. First, clinically feasible acquisition is sufficient for accurate reconstruction of quantitative measurements of microstructural properties in brain tissue of unknown orientation. Second, qualitative visualization using 3D interaction combined with anatomical landmarks can enhance user performance in isolating tracts for pathological analysis.

I focus my talk on the microstructural analysis of brain white matter, and briefly describe the visualization and interaction techniques at the beginning of the talk. The microstructural properties of brain white matter, such as axon radius, directly affect brain nerve functions and correlate with various neurological diseases. Despite their importance, these microstructural properties have not yet been measured reliably in vivo and their investigation has relied largely upon invasive histology examinations. At the microstructural level, existing diffusion-MRI measurements are nonspecific: they cannot distinguish among different sorts of microstructural changes that are of clinical importance. I present new computational approaches toward the virtual histology of brain tissue that reveal quantitative local measures of their microstructure and attempt to provide reliable and sensitive biomarkers for neurological changes. I have developed analytical models of water diffusion and incorporate them into a computational algorithm in order to extract underlying microstructural properties otherwise unattainable in vivo. I go beyond current experimental limitations, which require high gradients (not achievable in clinical scanners)and known tissue orientation, by using double-pulsed field gradient (d-PFG) diffusion MRI. I demonstrate that clinically feasible acquisition is sufficient to reconstruct these microstructural properties accurately in brain tissue of unknown orientation. I first quantitatively validate the feasibility and reliability of these computational approaches using simulations, which let us go beyond previous work to provide a complete study combining a set of characteristics that other work has used but never unified. For further validation of my methods, I then apply my computational approaches to live human subjects using clinical 3T MRI scanners for microstructure quantification and cross-subject comparisons. The results demonstrate the sensitivity of my approach to human microstructural properties, and also verify microstructural variations known from histology along the corpus callosum across different subjects.

Host: David Laidlaw