Template-based Automatic DTI Fiber Bundle Labeling

Introduction – We developed a template-based method for automatic labeling of white matter fiber bundles within whole-brain tractography models derived from diffusion-tensor imaging (DTI). The results show that the automatic labeling algorithm is able to correctly label six specific fiber bundles in two new subjects. We first ran automatic fiber clustering on DTI datasets of four healthy control subjects, then developed a template by manually labeling fiber-bundles-of-interest according to their shapes and locations, and finally automatically labeled the same fiber bundles from two new subjects using this template. This method aims to provide accurate white-matter tracts-of-interest (TOIs) for quantitative assessment of white matter. DTI tractography methods trace sufficiently large coherent neural pathways based on diffusion characteristics in cerebral white matter [1]. These pathways can be grouped into spatially related bundles that appear to represent underlying white matter anatomy. These DTI fiber bundles provide a potentially important research tool for improving our understanding white matter development or injury and its cognitive and behavioral correlates. However, to realize this potential, efficient and accurate automated methods are needed to reliably cluster, match, and label TOIs across individuals. We present an initial attempt at automatically matching and labeling the fiber bundles.