Automatic Tractography Segmentation by Morphological Continuity Clustering

Introduction Several tractography segmentation methods have been proposed to cluster the fibers tracts [1-4]. Most of the methods are based on training data provided by experts [3][4], which did not include short or unknown fiber tracts. Another clustering method depends on the assumption of the cluster number [2]. The cluster results may not be correct if the assumption is violated. In this study, we developed a fully automatic method that is unsupervised and without predefining the number of the clusters. It is based on the concept that the fiber tracts in the same cluster share the morphological continuity, which results in a continuous surface of the fiber bundle. This morphological continuity condition is used to construct the clusters that segment the whole brain fiber tracts. Theory Based on the concept of morphological continuity, we claim that for any two fiber trajectory ti(ri) and tj(rj), they belong to the same cluster if the following morphological continuity condition is satisfied: