Visual exploration of HARDI fibers with probabilistic tracking

Abstract

High angular resolution diffusion imaging (HARDI) is an effective method for characterizing complex neural fiber paths in the human brain. However, visualizing and analyzing the fibers is often challenging because of the complexity of the fiber orientation distribution function used to describe the crossing, kissing, and fanning fibers. In this paper, we propose a novel visual analytics approach to study brain fiber paths that allows users to explore fiber bundles to reveal the probability of fiber paths using a new visual classification method. First, we use a spherical deconvolution model for diffusion estimation and a Bayesian theorem for fiber tractography. Second, each fiber is subjected to a cluster analysis using pixel-based visual encoding. This result is shown in a pixel-based visual representation where each pixel bar maps opacity, color, and length to the probability, direction, and length of a fiber. Fiber bundles can then be acquired via a two-step classification routine that uses DBSCAN to group fibers based on similarities. Then the user can further refine fiber bundle selection using probabilistic information from the pixel bars. Therefore, the proposed approach shows not only the shape but also the confidence of the fiber paths. We demonstrate the resulting HARDI fiber bundles and compare a brain with tumor and normal brain using our system. Experiments and an empirical user study verify the effectiveness of our approach.

Uncertainty in the DTI Visualization Pipeline

“…Most of the Bayesian model-based techniques are often combined with random sampling methods, such as Markov Chain Monte Carlo (MCMC), to determine the distribution of model parameters [11,12,33]. The application of Bayesian model based methods in DTI and HARDI has been reported several times [48,54,64].…”

“…The PDF obtained from the analytical methods can be used to perform tractography with these stochastic techniques [33,48]. These studies are based on DTI, however, the concept is extendable to HARDI as well, but they are not used much in this context [64].…”