MODELING NEUROIMAGING DATA

Neuroimaging data have been analyzed in the past primary with inferential or exploratory statistical methods. We have been applying a pattern-based classification analysis to functional neuroimaging data of face and object perception from Haxby et. al. (2001). We begin by encoding fMRI brain scans in a principal component based space, as their projection coordinates. The PCA coordinates are then categorized using a linear discriminant classifier. The result is in the form of a discrimination matrix for the object categories. A second component of this work is to analyse the discriminability of the "stimuli" in conjunction with the discriminability of the brain maps. for more details....see poster.