Prefrontal-Hippocampal Interactions during Model-Based Learning

项目摘要 前额叶皮层（PFC）和海马（HPC）的功能障碍与许多 神经精神障碍，包括精神分裂症、重性抑郁症和创伤后应激障碍。许多 这些疾病的行为症状可以被建模为功能失调的强化学习（RL） 流程.例如，未能最佳地平衡目标导向（“基于模型”，或MB）和习惯性 （“无模型”，或MF）控制可以解释强迫症的反刍。我们研究的首要目标是 未来的设备发展将与神经回路以原则性的方式相互作用，以治疗神经精神疾病。 紊乱这种方法的一个障碍是，许多这些回路中的神经编码仍然很差 明白目前资助的目的是研究HPC和PFC的神经特性，以及如何 这些结构相互作用。 HPC-PFC电路可能在依赖于模型的基于奖励的学习过程中发挥重要作用 环境。HPC长期以来一直与表示“认知地图”相关联， 环境的结构。它与内侧前额叶皮层（mPFC）双向连接， 基于奖励的学习和基于价值的决策。我们的假设是，HPC-PFC是关键的MB RL，通过HPC表示任务的预测图，并将此图传达给mPFC以进行价值推断 引导行为的东西我们对这一过程建模的理论构造是后继表示（SR）， 学习任务的地图与奖励意外事件并行，然后通过整合来评估潜在的行动 预测地图与学习的奖励意外事件。 为了验证这一假设，我们开发了一个抽象的觅食任务，要求受试者在一个 隐藏的状态空间来寻找奖励。为了最佳地解决这一任务，受试者必须从事MB RL并开发 状态空间的内部映射。这张地图允许受试者存储最近一次奖励的位置 然后正确地选择到达奖励位置的必要动作。首先，我们将记录从单 在mPFC和HPC的神经元，然后同时记录mPFC和海马，以检查这些 区域在MB RL期间彼此通信。 总之，本提案的结果将扩大我们对HPC的作用和相互作用的理解 和前额叶皮层。这些知识不仅将为改善临床诊断工具的努力提供信息， 精神病学，但也可以奠定基础的神经假体设备的发展，将相互作用， 神经回路的原理来治疗神经精神疾病。

Project Summary Dysfunction of the prefrontal cortex (PFC) and the hippocampus (HPC) has been implicated in many neuropsychiatric disorders, including schizophrenia, major depression, and post-traumatic stress disorder. Many of the behavioral symptoms of these disorders can be modeled as dysfunctional reinforcement learning (RL) processes. For example, a failure to optimally balance goal-directed (“model-based”, or MB) and habitual (“model-free”, or MF) control can explain rumination in OCD. An overarching goal of our research is to inform the future development of devices that will interact with neural circuits in a principled way to treat neuropsychiatric disorders. One impediment to this approach is that the neural coding in many of these circuits remains poorly understood. The aim of the current grant is to investigate the neuronal properties of HPC and PFC, and how these structures interact with each other. The HPC-PFC circuit may play an important role for reward-based learning processes that depend on a model of the environment. The HPC has long been associated with representing a ‘cognitive map’ that encodes the structure of the environment. It is bidirectionally connected with medial PFC (mPFC), which has been implicated in reward-based learning and value-based decision-making. Our hypothesis is that the HPC-PFC is critical for MB RL, via the HPC representing a predictive map of task, and communicating this map to mPFC to allow value inferences that guide behavior. Our theoretical construct for modeling this process is the successor representation (SR), which learns a map of the task in parallel with reward contingencies, and then evaluates potential actions by integrating the predictive map with the learned reward contingencies. To test this hypothesis, we have developed an abstract foraging task that requires the subject to navigate a hidden state space to find a reward. To solve this task optimally, the subject must engage in MB RL and develop an internal map of the state space. This map allows the subject to store the location of the most recent reward and then correctly select the necessary actions to reach the rewarded location. First, we will record from single neurons in mPFC and HPC, then record simultaneously from mPFC and the hippocampus to examine how these regions communicate with each other during MB RL. Taken together, the results of this proposal will expand our understanding of the roles and interaction of HPC and PFC in the primate brain. This knowledge will not only inform efforts to improve diagnostic tools in clinical psychiatry but can also lay the groundwork for the development of neuroprosthetic devices that will interact with neural circuits in a principled way to treat neuropsychiatric disorders.