Institute for Data Analysis and Visualization

The structure of fiber tracts in DT-MRI data presents a challenging problem for visualization and analysis. We derive visualization of such traces from a local coherence measure and achieve much improved visual segmentation. We introduce a coherence measure defined for fiber tracts. This quantitative assessment is based on infinitesimal deviations of neighboring tracts and allows identification and segmentation of coherent fiber regions. We use a hardware-accelerated implementation to achieve interactive visualization on slices and provide several approaches to visualize coherence information. Furthermore, we enhance existing techniques by combining them with coherence. We demonstrate our method on both a canine heart, where the myocardial structure is visualized, and a human brain, where we achieve detailed visualization of major and minor fiber bundles in a quality similar to and exceeding fiber clustering approaches. Our approach allows detailed and fast visualization of important anatomical structures in DT-MRI data sets.