Fast parametrized volumetric DTI tract parcellation

A fast method is proposed for DTI neural tract extraction and visualization. Existing efficient tract extraction approaches are based on localizing fibers from a pre-computed whole brain tractography. However, the tracking parameters of precomputed fibers cannot be easily adjusted quickly. Moreover, high noise in the estimated orientation around crossing and neighboring tracts often diverts fibers away from desired pathways. Therefore, insufficient numbers of localized fibers decreases the quality of the tract visualization and also makes surface reconstruction difficult. Our method constructs a tract surface by considering tensor orientation along and around a single or double-seeded streamline and quickly expands it to a parametric surface. A noise reduction surface filter is applied to the parametric surface to provide a smooth surface rendering. Our method is particularly useful in areas of dense tracts such as the brainstem.