Statistical modeling and EM clustering of white matter fiber tracts

A statistical model of the fiber bundles is calculated as the average and standard deviation of a parametric representation of the fiber tracts, using the coefficients of the 3D quintic B-spline representation of the tracts. An atlas of the fiber tracts is constructed by averaging the bundle models over a population with the fiber tracts mapped onto the atlas coordinate. Using the model representation and with the atlas as the prior map, expectation-maximization (EM) is performed to cluster the fiber tracts in a mixture model framework. As an application, the method is applied to cluster the corpus callosum fiber tracts into its subdivisions and to calculate quantitative parameters for each region.