Analysis and visualization methods for interpretation of diffusion MRI data

Diffusion MRI is an imaging technique that is very sensitive to microstructural changes in tissue. Diffusion tensor MRI, the most commonly used method, can estimate the magnitude and anisotropy of diffusion. These tensor-based diffusion parameters have been shown to change in many neuropathological conditions, and are therefore a very sensitive imaging biomarker. Ever since the first in vivo MR measurements of anisotropic water diffusion in neuronal tissue, diffusion MRI is seen as having a great potential to investigate brain development, aging, and pathologies. To fulfill this potential, the entire diffusion MRI pipeline must be considered critically –from defining the research question, through data acquisition, analysis methods, and statistics to interpretation of results. The aim of this thesis was to improve this process, to make diffusion MRI a more sensitive and specific research tool. In this thesis I describe the influence of complex fiber architecture, e.g., crossing or bending fibers, on quantitative measures derived from DTI, including the fractional anisotropy and the mean diffusivity. These results demonstrate that the interpretation of DTI data is confounded by macrostructural organization, such as the size and shape of fiber bundles. Inclusion of such factors in the final analysis is required to make accurate inferences on changes in tissue microstructure. Diffusion MRI data can also be used to provide reconstructions of fiber pathways, called fiber tractography. The accuracy of these tractography reconstructions is strongly increased by the introduction of multi-fiber methods. Two important factors in the accuracy of fiber tractography are: i) how well can the intra-voxel fiber complexity be modeled (angular resolution), and ii) how well can different fiber populations be discriminated based on spatial location (spatial resolution). Ideally the data would have a higher spatial and a high angular resolution, but both come at the cost of extra scan time and a trade-off between these two factors is often made during acquisition. I have investigated the trade-off between having either a high angular resolution (up to 100 diffusion-weighting directions) or a high spatial resolution (up to 1 mm isotropic resolution) to determine what would be the most beneficial in cases of limited scan times. For tractography purposes, our results show that the largest gain is achieved by having a higher angular resolution that allows for a better fit of intra-voxel fiber crossings. Lastly, new visualization methods for these tractography results are presented. These techniques can visualize the local microstructure along the reconstructed fiber tract pathways, giving a more complete representation of the tissue organization