Sparse logistic regression for estimating time-varying functional connectivity networks: A simulation study

Understanding dynamics of the brain networks using functional Magnetic Resonance Imaging (fMRI) is an interesting neuroimaging topic, and sliding window based methods are widely employed to explore the time-varying functional connectivity networks. This paper presents a novel method for estimating dynamic functional connectivity networks using a Sparse Logistic Regression (SLR) model. This method uses a new protocol to optimize the problem of estimating time-varying networks based on a neighborhood selection procedure. To investigate the performance of the proposed approach, we have compared three time varying functional connectivity estimation techniques: combination of Sliding Window and Graphical Lasso (SWGL); Smooth Incremental Graphical Lasso Estimation (SINGLE), and SLR, in terms of their ability to recover true dynamics of functional connectivity networks. Simulation results demonstrate superiority of the SLR approach relative to the other two methods.