Visually Exploring Differences of DTI Fiber Models

Fiber tracking of Diffusion Tensor Imaging (DTI) datasets is a non-invasive tool to study the underlying fibrous structures in living tissues. However, DTI fibers may vary from subject to subject due to variations in anatomy, motions in scanning, and signal noise. In addition, fiber tracking parameters have a great influence on tracking results. Subtle changes of parameters can produce significantly different DTI fibers. Interactive exploration and analysis of differences among DTI fiber models are critical for the purposes of group comparison, atlas construction, and uncertainty analysis. Conventional approaches illustrate differences in the 3D space with either voxel-wise or fiber-based comparisons. Unfortunately, these approaches require an accurate alignment process and might give rise to visual clutter. This paper introduces a two-phase projection technique to reformulate a complex 3D fiber model as a unique 2D map for feature characterization and comparative analysis. To facilitate investigation, regions of significant differences among the 2D maps are further identified. Using these maps, differences that are difficult to be distinguished in the 3D space due to depth occlusion can be easily discovered. We design a visual exploration interface to study differences from multiple perspectives. We evaluate the effectiveness of our approach by examining two datasets.