Atlas-guided probabilistic diffusion-tensor fiber tractography

We demonstrate the use of a diffusion tensor atlas to perform probabilistic tractography in diffusion tensor data. The tensors from eleven subjects are normalized into a common space by optimizing the similarity between tensors explicitly. The distribution of tensor orientations in the atlas space forms a prior distribution for the fiber orientation that is defined by the local anatomy, in contrast to previous curvature priors that restrict the local curvature of all tracts equally. We demonstrate the method in a single subject and compare tracking with a uniform prior, with a curvature prior, and with the atlas prior. The atlas information allows us to track further along the fornix and cingulum than the other priors.