Resting-state functional magnetic resonance imaging (rs-fMRI) studies have focused primarily on characterizing functional or effective connectivity of discrete brain regions. A major drawback of this approach is that it does not provide a mechanistic understanding of brain cognitive function or dysfunction at cellular and circuit levels. To overcome this limitation, we combined the methods of computational neuroscience with traditional macroscale connectomic analysis and developed a Multiscale Neural Model Inversion (MNMI) framework that links microscale circuit interaction with macroscale network dynamics and estimates both local coupling and inter-regional connections via stochastic optimization based on blood oxygen-level dependent (BOLD) rs-fMRI. We applied this method to the rs-fMRI data of 66 normal healthy subjects and 66 individuals with major depressive disorder (MDD) to identify potential biomarkers at both local circuit and global network level. Results suggest that the recurrent excitation and inhibition within the dorsal lateral prefrontal cortex (dlPFC) might be disrupted in MDD, consistent with the commonly accepted hypothetical model of MDD. In addition, recurrent excitation in the thalamus was found to be abnormally elevated, which may be responsible to abnormal thalamocortical oscillations often observed in MDD. Overall, our modeling approach holds the promise to overcome the limitation of traditional large-scale connectome modeling by providing hidden mechanistic insights into neuroanatomy, circuit dynamics and pathophysiology.
Keywords: Functional connectivity; Major depressive disorder; Neural mass model; Resting-state fMRI; Stochastic optimization.