Functional independent components: revealing cortico-cortical, cross-frequency interactions

In its most classical and simple setting, the statistical analysis of functional brain images consists of comparing one group of subjects under two different conditions, or comparing two independent groups of subjects, using voxel-by-voxel statistics with correction for multiple testing. This approach is widely used in almost all brain imaging literature, where the aim is the “localization of brain function”. An alternative approach consists of abandoning the concept that brain function can be localized to a small number of hotspots, and to embrace the notion that brain function is more adequately characterized by the spatio-temporal/frequency distributions of activity. Recently, this concept is becoming popular with the use of independent component analysis (ICA) of fMRI data. In this paper, using insights from a novel field in statistics called “functional data analysis”, a new formulation for the discovery of interaction patterns of brain activity across space, time and frequency is presented. The method is illustrated by comparing EEG data between schizophrenic patients and normal control subjects.