White Matter Connectivity

Diffusion MRI offers microstructural information that is not available with conventional MRI modalities. Brain white matter fibers are often organized into colinear bundles at the scale of MRI resolution (1-2 mm), which makes it possible to perform tractography: tracing the path of fiber bundles and study the structural connectivity in vivo.

PICSL is advancing the state of the art in diffusion imaging by providing new methodologies and tools for the study of white matter structures. Diffusion data from multiple individuals is mapped to a population-specific template so that all subjects may be analyzed in a common space. Novel mathematical models of the white matter structures are used to perform quantitative analysis of the diffusion properties, which are sensitive to the microstructural architecture of the white matter.

Probabilistic tractography

Probabilistic tractography methods consider the uncertainty in the fiber orientation as part of the tracking algorithm and generate many possible paths that connect to a given starting point. We use a population-specific atlas to provide additional information to the probabilistic tracking algorithm. The atlas-guided tracking algorithm considers both the noise on the diffusion-weighted data, and the variability in fiber architecture present in the atlas. The atlas helps us to track fibers that are difficult to capture using the subject data alone, such as the fornix and cingulum.

Atlas-guided tracking The cingulum bundle is difficult to track using conventional fiber-tracking algorithms because of its curved, thin shape. Starting from the same point highlighted by the green arrow, we track the cingulum using a standard probabilistic tractography algorithm (top). Restricting the curvature of the fibers does not make much difference (middle) but using the atlas information allows us to track further along the fiber path (bottom).

Quantitative analysis of connectivity

We are exploring novel methods to compare fiber connectivity across populations, combining atlas building, fiber tractography and multivariate statistics. Click on the images for further information.

Methods for modeling white matter fiber structures (left) Model connections are visualized as tubes with colors corresponding to their cortical regions. The streamlines used to determine the model connections are shown with complementary colors. (right) Several major fiber tracts are sheet-like in appearance and can be modeled by a surface. Quantities of interest, such as mean diffusivity, can be smoothed on the surface and the surfaces can be flattened to provide an informative 2D visualization.

publications

7. (Online at Springer-Verlag) (pdf)