An introduction to anatomical ROI-based fMRI classification analysis

Modern cognitive neuroscience often thinks at the interface between anatomy and function, hypothesizing that one structure is important for a task while another is not. A flexible and sensitive way to test such hypotheses is to evaluate the pattern of activity in the specific structures using multivariate classification techniques. These methods consider the activation patterns across groups of voxels, and so are consistent with current theories of how information is encoded in the brain: that the pattern of activity in brain areas is more important than the activity of single neurons or voxels. Classification techniques can identify many types of activation patterns, and patterns unique to each subject or shared across subjects. This paper is an introduction to applying classification methods to functional magnetic resonance imaging (fMRI) data, particularly for region of interest (ROI) based hypotheses. The first section describes the main steps required for such analyses while the second illustrates these steps using a simple example.