Improved Probabilistic Tractography Using Atlas-Based Fiber Tracking

Qing Xu¹, Adam W. Anderson¹, John C. Gore¹, Zhaohua Ding¹

A probabilistic fiber tracking algorithm generates a set of fibers that reflect the distribution of underlying neuronal pathways by processing the diffusion tensor images. As imaging artifacts such as random noise and partial volume averaging usually render reconstructed fibers unreliable, typically certain prior knowledge is used to regularize the fiber tracking process. In this contribution, we propose a novel atlas based probabilistic fiber tracking algorithm that incorporates prior knowledge from a white matter fiber atlas, thereby eliminating the need for the commonly used heuristic priors.Preliminary experiments demonstrate that the atlas-guided method improves probabilistic tractography over methods with heuristic priors.