Feature Extraction for DW-MRI Visualization: The State of the Art and Beyond

By measuring the anisotropic self-diffusion rates of water, Diffusion Weighted Magnetic Resonance Imaging (DW-MRI) provides a unique noninvasive probe of fibrous tissue. In particular, it has been explored widely for imaging nerve fiber tracts in the human brain. Geometric features provide a quick visual overview of the complex datasets that arise from DW-MRI. At the same time, they build a bridge towards quantitative analysis, by extracting explicit representations of structures in the data that are relevant to specific research questions. Therefore, features in DWMRI data are an active research topic not only within scientific visualization, but have received considerable interest from the medical image analysis, neuroimaging, and computer vision communities. It is the goal of this paper to survey contributions from all these fields, concentrating on streamline clustering, edge detection and segmentation, topological methods, and extraction of anisotropy creases. We point out interrelations between these topics and make suggestions for future research.