Effects of regularizaron constraints on FMRI brain network source decomposition

Human brain functional connectivity can be reliably studied with the aid of 1MRI technology. Brain functional network decomposition can be solved by available methods such as Independent Component Analysis (ICA) with independence constraint, Morphological Component Analysis with KSVD dictionary update (MCA-KSVD) with sparsity constraint on spatial components, or constraint-free method PowerFactorization (PF) that has not been applied and known to the fMRI community so far. In the quest for finding methods that are effective for analyzing 1MRI functional networks, this study investigates the effects of various constraints used in the ICA MCA-KSVD and PF methods on the resulting decomposed networks. The observed mutual effects of independence and extreme sparsity constraints experimentally suggest that there is a connection between the two constraints. Specifically, the sparsity constraint in extreme case yields spatially independent components.