Reward Learning in Late-Life Suicidal Behavior

描述（由申请人提供）：老年人自杀脆弱性的神经机制在很大程度上仍然未知，部分原因是缺乏实证研究。尽管尸检和成像研究已经产生了一些候选的神经标记物，但它们与自杀的心理和认知途径无关。我们的实验室试图通过调查导致自杀行为的决策过程来弥合这一差距。自杀，尤其是老年自杀，通常被视为一种战略性的、理性的决定。相反，我们认为，自杀性危机往往会意外升级，灾难性的选择是由于未能适当地将瞬间的经验与预先存在的知识和价值观结合起来。从经验上看，这种错误的整合见于有自杀未遂史的人，他们倾向于根据基于奖励/惩罚的学习实验中的即时反馈做出错误的选择。申请人的K23支持的结合功能磁共振成像和计算建模的初步研究发现，自杀者未能整合积累的反馈是由中断的边缘预期奖励信号引起的。拟议的研究更进一步，通过在多个层面上检查自杀者的动态决策过程：通过BOLD激活测量的奖励信号，通过功能连接测量的皮质纹状体丘脑网络完整性，奖励学习期间的行为，决策中涉及的认知表现维度，以及与现实世界决策相关的心理特征。我们的第一个目标是研究皮质边缘奖励系统的改变， 在自杀行为中做出有偏见的决定。我们关注决策中重要信息的忽视，测试其与自杀未遂的关系，并研究其神经基础。我们的第二个目标是探索执行控制不良和自杀行为之间的联系。我们将测试一个假定的途径，从破坏新皮层自上而下的控制决策失败的行为适应自杀意念。我们提出了一个横断面病例对照研究，有三组抑郁症的参与者（30自杀未遂者，30自杀意念者，30非自杀抑郁症老年人）和30非精神病对照组，从一个正在进行的研究晚年自杀行为（PI：Szanto）招募。参与者将完成一个新的概率反向学习fMRI任务，该任务通过实验操作多个理论预测信号（预期奖励，奖励概率，奖励预测误差，波动性）。神经数据的分析将使用强化学习模型。抽样策略和详细评估旨在最大限度地减少和控制混杂因素。调查小组包括功能（Siegle）和特别是老年（Aizeinstein）成像，神经数据统计分析（Kass），决策神经科学（Clark）和认知控制（MacDonald）方面的专家。这项工作利用了形式学习理论和决策神经科学的强大理论框架。这将为自杀行为的维度分类和干预目标的发现奠定基础。

DESCRIPTION (provided by applicant): The neural mechanisms of vulnerability to suicide in old age remain largely unknown, in part due to a paucity of empirical research. Although post-mortem and imaging studies have yielded a number of candidate neural markers, they are not connected to psychological and cognitive pathways to suicide. Our lab attempts to bridge this gap by investigating decision processes that lead to suicidal behavior. Suicide, especially in old age, is often viewed as a strategic, rational decision. In contrast, we argue that the suicidal criis often escalates unexpectedly, and catastrophic choices result from a failure to appropriately integrate moment-to- moment experiences with pre-existing knowledge and values. Empirically, such faulty integration is seen in people with a history of attempted suicide, who tend to make poor choices based on moment-to-moment feedback in reward/punishment-based learning experiments. The applicant's K23-supported preliminary study combining fMRI and computational modeling found that the suicidal persons' failure to integrate accumulating feedback is paralleled by disrupted paralimbic expected reward signals. The proposed study goes a step further by examining dynamic decision processes in suicidal people at multiple levels: reward signals measured through BOLD activation, corticostriatothalamic network integrity measured by functional connectivity, behavior during reward learning, dimensions of cognitive performance implicated in decision- making, and psychological traits relevant to real-world decisions. Our first aim is to investigate corticolimbic reward system alterations underlying biased decision-making in suicidal behavior. We focus on the neglect of vital information in decisions, test its association with attempted suicide and examine its neural underpinnings. Our second aim explores the well-established link between poor executive control and suicidal behavior. We will test a putative pathway from disrupted neocortical top-down control over decision-making to failure of behavioral adaptation to suicidal ideation. We propose a cross-sectional case-control study with three groups of depressed participants (30 suicide attempters, 30 suicide ideators, 30 non-suicidal depressed elderly) and 30 non-psychiatric controls, recruited from an ongoing study of late-life suicidal behavior (PI: Szanto). Participants will complete a novel probabilistic reversal learning fMRI task, which experimentally manipulates multiple theoretically predicted signals (expected reward, reward probability, reward prediction error, volatility). Analyses of neural data will use a reinforcement learning model. The sampling strategy and detailed assessments are designed to minimize and control for confounding factors. The investigative team includes experts in functional (Siegle) and specifically geriatric (Aizeinstein) imaging, statistical analysis of neural data (Kass), decision neuroscience (Clark), and cognitive control (MacDonald). This work leverages powerful theoretical frameworks of formal learning theory and decision neuroscience. It will lay the groundwork for a dimensional classification of suicidal behavior and for discovery of intervention targets.