|
Activity Number:
|
190
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #302551 |
|
Title:
|
Bayesian Nonparametric Model for fMRI of the Visual Cortex
|
|
Author(s):
|
Raymond G. Hoffmann*+ and Nicholas M. Pajewski and Edward A. DeYoe and Daniel B. Rowe
|
|
Companies:
|
Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin
|
|
Address:
|
Division of Biostatistics, Milwaukee, WI, 53226,
|
|
Keywords:
|
non-parametric ; point process ; Dirichlet prior ; fMRI ; imaging ; spatial density
|
|
Abstract:
|
The fMRI visual field map (VFM) is obtained by using a rotating-expanding visual target to identify the area of the retina that corresponds to activated visual cortex. Since fMRI data is (1) relative rather than absolute and (2) has a degree of noise that may mask the activation, identifying differences in VFMs requires a model that will differentiate changes in the underlying structure from differences due to imaging variability. The VFM produces a non-homogenous, non-isotropic set of points on a disk that includes irregular features like the blind spot. A non-parametric mixture model, using a Dirichlet prior on a space of 2D density functions, will be used to model the VFM under experimental conditions where part of the visual field is masked by a circular wedge. The posterior probability of the difference in the models, will be used to quantify the probable location of the wedge.
|
