|
Activity Number:
|
271
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 1, 2016 : 3:05 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #321820
|
|
|
Title:
|
Large-Scale MCMC Using GPU with Application in Brain Imaging
|
|
Author(s):
|
Yang Yang* and Galin Jones
|
|
Companies:
|
University of Minnesota and University of Minnesota - Twin Cities
|
|
Keywords:
|
MCMC ;
GPU ;
Brain Imaging ;
Bayesian ;
Computing ;
MRI
|
|
Abstract:
|
We present a framework to assess the uncertainty in neuron fiber orientations estimation via diffusion magnetic resonance imaging (dMRI). A commonly used model in dMRI is a simple partial volume model of diffusion. This results in a Bayesian model whose parameters define local neuron fiber direction. Markov Chain Monte Carlo (MCMC) is implemented in the estimation process. We develop a rigorous method for ascertaining how many Monte Carlo samples are required to ensure reliable estimation. We apply our method to an example and show that the Monte Carlo sample sizes typically used in applications are too small due to computational limits. To demonstrate the scalability of our approach, we implement GPUs to apply our method to human brain data which consists of more than 3 million voxels. We will also discuss the advantage and limitation of GPU computing.
|
Authors who are presenting talks have a * after their name.
