Multivariate Pattern Classification of fMRI data
Research Field:Life Sciences
Lead PI:Prof. John O'Docherty
Abstract:Multivariate fMRI studies is an emerging trend in cognitive neuroscience. Standard practice in neuroimaging is to analyses data on a voxel-by-voxel basis. Voxel responses are independently examined for significant increases in activity using general linear models (GLMs). Unfortunately, this approach is fundamentally limited in what it can achieve since researchers are restricted to inferring whether a particular region of the brain is significantly active or not during a particular cognitive process. In recent years, these limitations have begun to be overcome by the use of multivariate methods which make use of distributed patterns of neural activity. That is, these techniques elicit combinatorial codes of voxel values which can represent cognitive functions. Whereas univariate GLM analyses can highlight where in the brain significant neural activity occurs, multivariate methods can go beyond this to determine what information is being processed and begin to investigate how. The goal of the current project is to apply multivariate pattern analysis to fMRI data from studies on reward learning and decision making.