1

Abstract—MRI diffusion imaging of the human brain provides neuron fiber trajectories that can be grouped into different bundles that have distinctive neurological functions. In this paper, three major fiber bundles passing through brainstem are identified. 3D shape context is used to represent each fiber. Various manifold learning algorithms are applied to reduce the dimensionality of this histogram feature, including PCA, KPCA, Isomap, LLE, Laplacian eigenmaps. Supervised classification with K-Nearest neighborhood is carried out to examine the performance of different dimensionality reduction methods. A leave-one-out validation is done in 10 subjects. The results show the feasibility of the manifold learning technique to fiber bundling and the importance of dimensionality reduction in the classification problem. Index Terms—Diffusion tensor images; fiber classification; shape context; supervised learning; manifold learning I