Towards inference of human brain connectivity from MR diffusion tensor data

Abstract

This paper describes a method to infer the connectivity induced by white matter fibers in the living human brain. This method stems from magnetic resonance tensor imaging (DTI), a technique which gives access to fiber orientations. Given typical DTI spatial resolution, connectivity is addressed at the level of fascicles made up by a bunch of parallel fibers. We propose first an algorithm dedicated to fascicle tracking in a direction map inferred from diffusion data. This algorithm takes into account fan-shaped fascicle forks usual in actual white matter organization. Then, we propose a method of inferring a regularized direction map from diffusion data in order to improve the robustness of the tracking. The regularization stems from an analogy between white matter organization and spaghetti plates. Finally, we propose a study of the tracking behavior according to the weight given to the regularization and some examples of the tracking results with in vivo human brain data.