Novel Approach for Modeling Event-Related fMRI Data.

Katie Gates, Penn State

February 15, 2010 @ 12:10 pm to 01:15 pm

106 Boucke

Functional MRI researchers increasingly wish to understand how the brainorchestrates responses. Most current approaches look at contemporaneous relations among regions to arrive at separate models for each block (e.g., resting state and task). While intuitive, this approach disregards information that could be vital to understanding brain functioning. First, like most biological data, BOLD activity contains temporal dependencies which may bias estimates. A second, related, issue is that the temporal ordering of regional activities may elucidate how the brain functions. Third, external stimuli may influence regional activity directly or indirectly via its impact on another region.__ Current methodologies do not attend to all of these concerns. We present a new technique, the extended unified SEM, for modeling fMRI data which attends to these issues. An added benefit of the present approach is that it may be used in a data-driven manner to arrive at a final model without prior specification of directional paths.