Diffusion Weighted MRI is a recent modality to investigate neuronal
pathways of the brain. In order to achieve understandable
visualizations of the resulting datasets, this dissertation reduces
them to relevant features. First, the accuracy of fiber tracking
methods in regions of crossing fibers is increased, and streamlines
are put into context with structural MR images. Then, derivative-based
methods are employed to identify boundaries, to segment meaningful
regions, and to describe local variance in the data. Finally, the role
of tensor topology for the visualization of diffusion tensors is
clarified.