Action-mode subnetworks for decision-making, action control, and feedback

Abstract

The action-mode network (AMN) is a canonical functional brain network first identified using resting-state functional connectivity (RSFC). Based on animal and human data, we have proposed that AMN supports the brain’s action mode by controlling functions required for successful goal-directed behavior. However, task fMRI averaged across groups has associated AMN regions with a variety of behaviors, contributing to uncertainty about AMN function. Here, we investigated the AMN using an inside-out approach, in which the network architecture of the AMN is first precisely mapped within individuals and then associated with behavioral functions. Individual-specific precision functional mapping with >5 h of RSFC and task functional magnetic resonance imaging (fMRI) data revealed a replicable AMN subnetwork structure. AMN subnetworks were characterized and annotated by combining a meta-analytic network association method with RSFC, intrinsic timing, and task activation profiling. We demonstrate the existence of AMN-Decision, -Action, and -Feedback subnetworks that are distributed across lobes, forming a temporally sequential within-network processing stream by which the brain adjudicates between possible goals, sets action plans, and modifies those plans in response to feedback such as pain. A subnetwork in the pars marginalis of the cingulate was distinct from the Decision, Action, and Feedback subnetworks and may be important for the construction of the bodily self.