Department of Statistics Seminar
North Carolina State University

presents

Dr. DuBois Bowman

Emory University

"Modeling Spatial Correlations in Functional Neuroimaging Data "

ABSTRACT

Functional neuroimaging studies attempt to determine spatially localized brain regions that drive the execution of particular experimental tasks targeting, for example, cognition, emotion, or behavior. These studies produce massive data sets comprised of serial scans on each subject, with each scan containing hundreds of thousands of spatially localized measurement sites (voxels). The data exhibit temporal correlations between repeated scans and complex patterns of spatial correlations between voxels. Temporal correlations are often addressed using variants of autoregressive models or other repeated measures covariance structures. Attempts to incorporate spatial correlations frequently involve pooling information from contiguous voxels in the brain. In this talk, we explore methods for modeling the spatial correlations in neuroimaging data that extend beyond first-order neighbors, while also capturing temporal correlations across repeated observations. We illustrate the use of our methods to determine distributed patterns of behavior-related changes in measured brain activity and to assess task-related functional connectivity within the brain.

Friday, November, 04, 2005

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.