Modelling noise-induced fibre-orientation error in diffusion-tensor MRI

Diffusion-Tensor MRI can be used to measure fibre orientation within the brain. Several studies have proposed methods to reconstruct known white matter fibre tracts in the brain. These methods are known as tractography. However, the measured fibre orientations are subject to error, which leads tractography methods to fail or define false connections. Probabilistic tractography methods use a model of the probability density function (PDF) of the local fibre orientation in each voxel, to calculate the likelihood of any potential fibre pathway through a DT data set. We propose the Watson distribution as a new fibre orientation PDF to replace ad hoc models used previously. We compare the Probabilistic Index of Connectivity (PICo) tractography method using three candidate PDFs and show that the Watson PDF compares favourably to the ad hoc models.