Estimation of Task-Evoked Dynamic Effective Connectivity from fMRI in A Visual Target Detection Oddball Experiment

Abstract

Studies of dynamic effective connectivity (dEC) are important to understand the causal relationships among different brain regions. This paper presents a novel method for fMRI data analysis by estimating dEC using a time-varying autoregressive (TVAR) model with spatial sparsity and temporal continuity constraints. The constrained TVAR model is solved as a least square problem with L 1 -regularizations for making inference of dEC. To effectively solve the resultant large-scale optimization problem, the derivation of the gradient is also provided so that effective optimizer can be used to reduce the computational complexity and memory requirement. The proposed method is then applied to fMRI data recorded in a visual target detection oddball experiment to estimate task-evoked dEC. To validate the fMRI dEC, we use graph theoretic measures to demonstrate the dynamically properties changes of various of brain regions between tasks. The statistical tests and the result analysis show that the method can effectively capture the brain connectivity patterns among different brain regions, thus verify the effectiveness of the proposed method and suggest it as a promising tool for investigating task-related dEC.