Hierarchical Poisson-Disk Sampling for Fiber Stipples

Abstract

To understand neural tracts of the brain, neuroscientists use visualizations of diffusion data. Fiber stippling - a technique inspired by illustrations - accommodates probabilistic tracts, main diffusion direction, and anatomical context in the same slice image. It uses stratified sampling to place stipples, but this can result in overlaps and undersampled areas that distort the perception of tract probability. Moreover, when changing slices in an interactive setting, resampling leads to visual noise that distracts from real changes in the data. In this paper, we propose to use Poisson-disk samplings to ensure adequate pattern perception inside slices and a hierarchy of samplings to ensure coherence among slices. We also port the algorithm to the GPU to achieve interactive frame rates. Our modifications are appreciated by neuroscientists, who can now investigate white-matter structures more confidently.