|
Abstract:
|
Diffusion Tensor Imaging is a popular procedure for identifying the anatomical structure of white matter in VIVO. Due to the ambiguity of directions in the cases of crossing fibers, the orientation statistics summarized as DTs may not represent the diffusion pattern perfectly. As a result, the deterministic tractography to construct the possible fiber tracts may not uncover all structural connectivity pattern.The probabilistic tractography provides a way to explore a wide range of fiber tracts while the procedure can be slow in the case of higher resolution data. In this article, we present a novel Bayesian procedure that starts with clustering of the directional data using a mixture of Von-Mises-Fisher distribution to summarize high-resolution data and eventually perform tractography on the summarized data. Specifically, for the parameter estimation, we develop an efficient MCMC algorithm as well as a faster alternative based on optimization. Our model consists of tuning parameters which could be used to adjust the quality of summarization as per the need of the specific condition. We validate the proposed techniques with appropriate simulations and real DTI data analysis.
|
