Date: September, 22nd, Thursday
Time: 3:00 pm
Place: Little 339 (the Atrium)
Speaker: Weihong Guo

Title:

Markov chain monte carlo based white matter fiber tracking in
diffusion weighted magnetic resonance imaging

Abstract:

Fiber tracking is the only noninvasive technique that can
visualize the neuronal connection between functional regions in
the human brain. Due to noise and artifacts present in any
magnetic resonance(MR) scan and incomplete modeling of the signal,
there is always uncertainty about any information derived from the
MR data. Hence, it does not make too much sense to derive
determined result from noisy data. Markov chain monte carlo(MCMC)
based fiber tracking method is currently believed to be the most
senseful and convincing scheme as it provides probabilistic
result.

In this talk, I will introduce how to use Metropolis-Hastings
algorithm to sample from posterior distribution of unknown
parameters in diffusion tensor model. Idea of MCMC is then applied
to characterize uncertainty of each parameter and probability of
existing a fiber path between two given functional regions in
human brain.