Over-Complete Analysis for Resting-State fMRI Data

Abstract

Independent component analysis (ICA) has been widely applied to reveal brain networks from fMRI data. Because the ICA method is based on a complete linear model, it may not work well for the fMRI data with very few time points. Lewicki and Sejnowski (Neural Comput 12:337–365, 2000 [2]) proposed an algorithm for learning an over-complete basis. However, the over-complete model has not been applied to the fMRI data analysis. In this study, we investigate the feasibility of the over-complete analysis method in resting fMRI data. Our results demonstrate that the resting brain networks can be obtained from the resting fMRI data with short time points via the over-complete analysis method.