[life SCIENCES]

F unctional magnetic resonance imaging (fMRI) is a noninvasive, powerful tool that has been utilized in both research and clinical arenas since the early 1990s [1] and has provided valuable insights to the understanding of the human brain function. fMRI has enabled researchers to directly study the temporal and spatial changes in the brain as a function of various stimuli. Because it relies on the detection of small intensity changes over time, fMRI poses significant challenges for data analysis techniques. Traditional model-based analysis approaches—such as linear regression— are robust, yet often too rigid to capture the richness of the human brain activation. Independent component analysis (ICA), on the other hand, is a datacentric approach that provides a more flexible framework for the analysis of fMRI data. For simplicity and tractability, most fMRI analysis techniques to date have discarded the phase of the fMRI data. However, the phase information may be quite valuable for the analysis of the natively complex fMRI data. ICA facilitates the analysis of fMRI data in its complex form by eliminating the need to explicitly model the phase behavior. In what follows, we discuss the application of real and complex-valued ICA to fMRI data analysis and present two examples where ICA has proved particularly useful.