Contributive sources analysis: A measure of neural networks' contribution to brain activations

In this paper, we propose a new supplementary method that can give more insight into the functional data as well as help to clarify inconsistencies between the results of studies using GLM and ICA. We introduce a contributive sources analysis (CSA), which provides a measure of the number and the strength of the neural networks that significantly contribute to brain activation. CSA, applied to fMRI data of anti-saccades, enabled us to verify whether the brain regions involved in the task are dominated by a single network or serve as key nodes for particular networks interaction. Moreover, when applying CSA to the atlas-defined regions-of-interest, results indicated that activity of the parieto-medial temporal network was suppressed by the eye field network and the default mode network. Thus, this effect of networks cancelation explains the absence of parieto-medial temporal activation within the GLM results. Together, those findings indicate that brain activations are a result of complex network interactions. Applying CSA appears to be a useful tool to reveal additional findings outside the scope of the “fixed-model” GLM and data-driven ICA approaches.