An exploratory approach to modelling effective connectivity

Previously, sequential equation modelling (SEM) was applied to compute interactions of neural system derived from fMRI, which was understood as confirmatory modelling. In this paper, a data-driven approach with SEM is proposed to explore effective connectivity among brain units. Spatial ICA (independent component analysis) technique is used as an exploratory detection method for activation regions, and is combined with an automated best model search to build a best-fitted path model. This work was demonstrated with the visual-task fMRI data. We conclude by suggesting that this exploratory method will give a more dynamic model for determining the interactions among functional brain network, especially in the event of brain abnormalities.