Influence of Temporal Difference Reward Prediction Errors on Brain Network Connectivity during Learning and Decision-Making

点击翻译按钮获取中文摘要

PROJECT SUMMARY Disorders of choice behavior, such as substance abuse and impulse control disorders, involve over-valuing and repeatedly choosing certain reinforcers (e.g., drugs, risky actions), even in the face of recurring consequences. Understanding the link between adaptive choice behaviors and underlying neural activity is a strategic focus for substance abuse and mental health research. Much empirical evidence demonstrates that features of distinct pathological behaviors map onto distinct patterns of interactions between distributed brain regions. Nevertheless, how adaptive learning signals concerning rewards and punishments alter region-to-region functional interactions in real-time lies at the limit of our current understanding. As such, we seek to identify neurobehavioral markers that reflect how reinforcement learning (RL) signals alter functional brain network interactions and associated choice behaviors in healthy adults. Along this line of inquiry, this proposal’s central objective is to understand how real-time changes in inter-regional functional interactions – between, for instance, regions of the basal ganglia and limbic, prefrontal, and sensorimotor cortices – in response to RL signals influence adaptive choice behaviors. Our approach uses computational methods to investigate the quantitative relationship between measures of human choice behavior and brain network interactions at high-resolution spatiotemporal scales. Specifically, we pair computational RL models of human behavior on a probabilistic reward and punishment learning task with multi-modal functional neuroimaging to investigate changes in functional brain networks responsive to learning signals called ‘temporal difference reward prediction errors’ (TD RPEs). We will identify functional network interactions related to TD RPE signals to address, through two Specific Aims, our overarching hypothesis that TD RPE signals alter – in real time – the coupling (synchrony) of functional interactions between brain regions involved in processing rewards and punishments to direct changes in choice behavior. In Aim 1, we will measure functional networks interactions using magnetoencephalography (MEG) to test the hypotheses that (1) positive TD RPE signals increase, in real time, the coupling of inter-regional functional interactions and (2) negative or zero TD RPE signals decrease this coupling. In Aim 2, we use functional magnetic resonance imaging (fMRI) – within the same subjects from Aim 1 – to investigate individual-specific functional networks associated with TD RPE signals. Here, we will localize TD RPE-responsive functional regions-of-interest (ROIs) and incorporate these ROIs as individual-specific spatial priors for within-subject MEG analysis of functional network interactions. In all, this fellowship will provide training in contemporary modeling methods and experimental design that address fundamental questions in computational neuroscience regarding the functional organization of the human brain during adaptive learning. The central findings of this research will advance our understanding of the principles of brain activity supporting adaptive learning processes that are characteristically the root cause of many behavioral symptoms of substance abuse and related behavioral disorders.