Decision Making Journal Club
Multivariate dynamical systems models for estimating causal interactions in fMRI
Context dependent dynamical causal interactions between distributed brain regions provide powerful insights in understanding brain function and dysfunction. Therefore developing and validating computational methods for investigating causal interactions has taken a great significance. However, estimating causal interactions from functional magnetic resonance imaging (fMRI) is challenging because (a) fMRI measures Blood Oxygenation Level Dependent (BOLD) signals, rather than the underlying latent neuronal activity and (b) regional variations in the hemodynamic response function (HRF) can significantly influence estimation of casual interactions between them. To address this critical gap, we developed a state-space multivariate dynamical systems (MDS) model to estimate causal interactions between distributed brain regions in fMRI. To validate the performance of MDS, we applied it on several simulated fMRI data sets generated from realistic models of neuronal signals. Moreover, we took a unique and exciting new approach using optogenetic fMRI (ofMRI) to further validate MDS in vivo. Optogenetic probes were used to selectively stimulate brain regions with an optical signal during high-resolution whole-brain fMRI acquisition. This technique allow us to characterize the causal effects of stimulation on fMRI signals in target regions, and is therefore ideal for validating dynamic causal estimation methods in vivo. Finally, we applied MDS on experimental human fMRI data to investigate the causal interactions in a fronto-parietal network during an n-back task. In this talk, I will present results of aforementioned series of investigations.