Spatial independent component analysis of multitask-related activation in fMRI data

Abstract

Independent component analysis (ICA) is a technique to separate the mixed signal into independent components without priori assumptions about the hemodynamic response to the task. Spatial ICA (SICA) is applied widely in fMRI data because the spatial dimension of fMRI data is larger than their temporal dimension. The general linear model (GLM) is based on a priori knowledge about stimulation paradigm. In our study, a two-task cognitive experiment was designed, and SICA and GLM were applied to analyze these fMRI data. Both methods could easily find some common areas activated by two tasks. However, SICA could also find more accurate areas activated by different single task in specific brain areas than GLM. The results demonstrate that ICA methodology can supply us more information or the intrinsic structure of the data especially when multitask-related components are presented in the data. © Springer-Verlag Berlin Heidelberg 2003.