|
Activity Number:
|
72
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305208 |
|
Title:
|
A Dirichlet Process Model for Changes in the fMRI Visual Field Map
|
|
Author(s):
|
Raymond G. Hoffmann*+ and Pippa Simpson and Daniel B. Rowe
|
|
Companies:
|
Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin
|
|
Address:
|
8701 Watertown Plank Rd, Milwaukee, WI, 53226,
|
|
Keywords:
|
fMRI ; Diriclet Process ; Bayesian Model ; Imaging ; Spatial Model
|
|
Abstract:
|
The Visual Field Map (VFM) is a circular region that maps the visual cortex to a virtual retina. The relationship between the dynamic image presented to the eye and the virtual retina can be used to identify changes resulting from surgery or disease. The VFM is a nonisotropic, nonhomogeneous set of points that represent the activation of the visual cortex assessed with fMRI. A wedge shaped mask (18 to 90 degrees of arc) of the image was used to simulate the effect of surgical damage. A Bayesian nonparametric mixture model, a Dependent Dirichlet Process (DDP), with a Dirichlet prior on a space of 2D density functions G was used to model the intensity of the stochastic process that generates the points in the VFM. The posterior probability of the DDP model on the disk quantifies the probable location of the wedge-shaped mask compared to a reference scan.
|
