Identifying dynamic reproducible brain states using a predictive modelling approach

Conceptually, brain states reflect some combination of the internal mental processes of a person, and the influence of their external environment. Importantly, for neuroimaging, brain states may impact brain-based modeling of a person's traits, which should be independent of moment-to-moment changes in behavior. Investigation of brain states, and modeling of traits or behaviors are both often done using fMRI-based functional connectivity. Brain states can fluctuate in time periods shorter than a typical fMRI scan, and an array of methods called dynamic functional connectivity analyses has been developed to measure them. It has previously been shown that brain state can be manipulated through the use of continuous performance tasks that put the brain in a particular configuration while the task is performed. Here, we focus on moment-to-moment changes in brain state and test the hypothesis that there are particular brain-states that maximize brain-trait modeling performance. We use a regression-based framework, Connectome-based Predictive Modelling, allied to a resample aggregating approach, to identify behavior and trait-related brain states, as represented by dynamic functional connectivity maps. We find that there is not a particular brain state that is optimal for trait-based prediction, and combining data from distinct brain states across the scan is better. We also find that this is not the case for in-scanner behavioral prediction where more isolated and temporally specific parts of the scan session are better for building predictive models of behavior. The resample aggregated dynamic functional connectivity models of behavior replicated in sample using unseen left-out data. The modeling framework also showed success in estimating variance in behavior in a separate dataset. The method detailed here may prove useful for both the study of behaviorally related brain states, and for short-time predictive modeling.