We have suggested a method for mapping the number of fibre orientations in each voxel of a 3D diffusion MRI acquisition. This information can be used for selecting the most appropriate algorithm for finding the orientations of fibres in each voxel. We present a computationally expensive and accurate model generation and selection algorithm. The approach described does have some limitations, as it does not allow for oblate fibre orientation distributions, which could be expected in various brain regions in the presence of fanning (e.g. corona radiata) and of high curvature (e.g. optic radiation).
We compare results from this new method with those obtained from an existing algorithm, and show that the new method provides denser clusters of fibre-crossing voxels, which suggests more reliable identification. The results generated from this method share similarities to those obtained using the method described by other authors, as can be seen by comparing the images shown above, and by comparing the distribution of isolated voxels. We might improve our method by using global optimization techniques. The new method is computationally expensive, but we plan to use the output as a ‘gold standard’ for simpler and faster ways of mapping the number of distinct fibre populations.