MSBLFFM: Multivariate Semiparametric Bayesian Local Factor Functional Models for Diffusion Tensor Tract Statistics

• Activity Number: 310
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Graphics
• Author(s): Zhaowei Hua*+ and David Dunson and Hongtu Zhu
• Companies: The University of North Carolina at Chapel Hill and Duke University and The University of North Carolina
• Keywords: Bayes confidence band ; Diffusion tensor imaging ; Fiber bundle ; local inference ; Dirichlet process

Abstract

Diffusion tensor imaging (DTI) is a modality to visualize and quantify the structure of white matters in human brain. In this article, we propose a multivariate semiparametric Bayesian local factor functional model to analyze fiber tract data. A local partition process is used to address the variability of multiple diffusion properties along major white fiber bundles and its association with a set of covariates of interests, such as gestational age. Two types of statistical inferences are provided: (1) global hypothesis testing to test the overall significance of a hypothesis of interest (2) local hypothesis testing to identify the region of significance. Posterior computation proceeds via an efficient MCMC algorithm using the exact block Gibbs sampler. A simulation study is performed to evaluate the performance of MSBLFFM. Our method is applied to analyze a fiber track data set of two fiber tracts from a clinical study of neurodevelopment: the splenium of the corpus callosum tract and the right internal capsule tract. The growth of white matter fiber diffusivities along these two tracts are addressed.