Neural Substrates of Reinforcement Learning and its Training in Major Depression

 描述（由申请人提供）：重度抑郁症是世界上死亡和残疾的最重要原因之一。来自我们小组和其他人的最新数据强调，奖励和损失学习障碍是抑郁症的核心，具有不同的神经基质，并通过成功的治疗得到改善。总之，这些发现表明迫切需要描绘神经行为奖励和损失，学习障碍和抑郁症之间的关系。同样重要的是，这些见解提出了新的干预目标，例如操纵奖励和损失学习的神经行为底物可能有助于抑郁症的症状改变。为了解决这些问题，我们使用功能神经成像和定量强化学习框架来i）系统地表征参与抑郁症中奖励和损失学习障碍的神经和行为基质（目标1），以及ii）评估这些障碍对两种直接针对奖励和损失学习的训练方法的反应程度（目标2和3）。在目标2和3中，我们利用了我们小组和其他人的现有数据，这些数据表明，外显和隐性任务修改分别导致控制中的自适应神经和行为学习变化。在这里，我们将这项工作扩展到抑郁症患者，并测试了广泛的假设，即i）抑郁症的特征可能是奖励学习和损失学习中不同的神经行为障碍，以及ii）这些缺陷可以通过有针对性的行为训练来正常化。基于计算模型的强化学习分析的最新进展提供了一个强大的神经机制框架，在该框架内描绘了建议的奖励和损失学习障碍及其改善的性质和轨迹。

 DESCRIPTION (provided by applicant): Major depressive disorder ranks among the most significant causes of mortality and disability in the world. Recent data from our group and others highlight that impairments in reward and loss learning are central to depression, have distinct neural substrates, and improve with successful treatment. Together, these findings suggest an urgent need to delineate the relationships among neurobehavioral reward and loss learning impairments and depression. Equally important, these insights suggest novel targets for intervention such that manipulating the neurobehavioral substrates of reward and loss learning may facilitate symptom change in depression. To address these issues, we use functional neuroimaging and a quantitative reinforcement learning framework to i) systematically characterize the neural and behavioral substrates that attend reward- and loss- learning impairments in depression (Aim 1), and ii) assess the degree to which these impairments respond to two methods of training that directly target reward and loss learning from different angles (Aims 2 and 3). In Aims 2 and 3, we capitalize on extant data from our group and others showing that explicit and covert task modifications, respectively, lead to adaptive neural and behavioral learning changes in controls. Here we extend this work to individuals with depression and test the broad hypotheses that i) that depression may be characterized by distinct neurobehavioral impairments in reward- and loss- learning, and ii) these deficits may be normalized through targeted behavioral training. Recent advances in computational model-based analyses of reinforcement learning provide a robust neuromechanistic framework within which to delineate the nature and trajectory of the suggested reward- and loss- learning impairments and their amelioration.