Assessing space and feature based attention with fMRI and multivoxel pattern anal

描述（由申请人提供）：感知是认知的基石；记忆、推理和运动计划的形成都依赖于对周围环境形成稳定表征的能力。然而，对感觉输入进行编码的神经元本质上是嘈杂的，所以一个刺激的重复呈现不会两次唤起相同的神经活动模式。此外，环境中存在的刺激通常比大脑同时处理的刺激多，因此相关和不相关的项目必须在皮层表征中竞争。考虑到这些障碍，大脑形成连贯感知的能力是相当了不起的，理解这一壮举是如何完成的，是揭示支持有意识意识的神经机制的重要的第一步。理论研究表明，由于平均可以减少噪声（称为种群编码模型），感知是基于汇总输出的小群体神经元的。此外，注意特定的位置或特征会使神经活动产生偏差，从而使相关刺激以牺牲无关刺激为代价赢得表征。因此，群体编码方案是必要的，为感知提供可靠的基础，并需要选择性注意，以确保相关刺激的表征主导意识。不幸的是，注意力和人口编码之间的关系并没有得到很好的理解，部分原因是技术上的限制，部分原因是很少有工作将单个单位的注意力调节与人口水平的信息编码效率联系起来。在这里，我们使用一个简单的计算模型来提供注意调节、人口编码和感知之间的明确联系。为了验证该模型产生的预测，我们结合了心理物理学和新型的多变量功能磁共振成像（fMRI）分析技术，这些技术对视觉皮层特征选择区域的人群反应谱变化非常敏感。具体来说，我们提出了一种在很小的早期视觉皮层区域内测量特征选择“调谐功能”的方法。这些基于fmri的调谐功能类似于非人类灵长类动物的单单元记录方法常规获得的调谐功能，为评估人类感觉皮质的信息编码理论提供了强大的工具。在第一个特定目标中，我们将使用这些工具集来测试不同的注意增益模型（例如乘法增益与对比增益），并确定基于特征的注意是否在出现刺激之前就系统性地影响了早期视觉皮层的总体反应曲线。在第二个目标中，我们批判性地评估了一种普遍的直觉，即注意力增益应该应用于对目标刺激反应最大的感觉神经元。相反，我们将测试反直觉的预测，增益有时应该应用于实际上没有调整到参与特征的神经元，以最大化人口代码的效率。总之，这些努力将阐明观察者的行为目标如何影响人群的反应概况，从而优化视觉系统内的信息处理。由于感知是人类信息处理的一个基本方面，我们的研究结果将引起关注感知学习、决策和记忆的研究者的兴趣。此外，本文所获得的知识将为今后注意缺陷障碍（attention Deficit Disorder， ADD）等注意力障碍的应用研究提供重要的基础。例如，更好地理解注意力如何调节感觉神经元可能会导致更客观的诊断测试，从而更早、更准确地识别这些疾病。无论是在教室里听老师讲课还是在路上开车，关注环境重要部分的能力对我们的成功和生存至关重要。在目前的研究计划中，我们试图了解大脑中的神经活动模式如何支持注意行为，以便观察者可以更有效地理解和代表传入的感觉信息。这一知识将有助于为常见的注意力障碍（如注意力缺陷障碍）开发更客观的测试，从而使诊断能够更精确地进行，并使干预措施能够更早地开始。

Description (provided by applicant): Perception is the cornerstone of cognition; memory, reasoning, and the formation of motor plans all rely on the ability to create a stable representation of the surrounding environment. However, neurons that encode sensory input are inherently noisy, so repeated presentations of a stimulus never evoke the same pattern of neural activity twice. In addition, more stimuli are typically present in the environment than the brain can simultaneously process, so relevant and irrelevant items must compete for cortical representation. Given these obstacles, the brain's ability to form coherent percepts is quite remarkable and understanding how this feat is accomplished is an important first step towards revealing the neural mechanisms that support conscious awareness. Theoretical studies suggest that perception is based on small populations of neurons that pool their output, since averaging reduces noise (termed population coding models). In addition, attending to specific locations or features biases neural activity so that relevant stimuli win representation at the expense of irrelevant stimuli. Thus, population coding schemes are necessary to provide a reliable foundation for perception, and selective attention is required to ensure that representations of relevant stimuli dominate awareness. Unfortunately, the relationship between attention and population codes is not well understood, in part because of technical limitations and in part because little work has been done to link single-unit attention modulations with the efficiency of information encoding at the population level. Here, we use a simple computational model to provide an explicit link between attention modulations, population codes and perception. To test predictions generated by the model, we use a combination of psychophysics and novel multivariate functional magnetic resonance imaging (fMRI) analysis techniques that are sensitive to changes in population response profiles across feature-selective regions of visual cortex. Specifically, we present a method for measuring feature-selective `tuning-functions' within very small regions of early visual cortex. These fMRI-based tuning functions resemble the tuning functions routinely obtained using single-unit recording methods in non-human primates, providing a powerful tool for evaluating theories of information encoding in human sensory cortices. In the first Specific Aim, we will use this collection of tools to test different models of attention gain (e.g. multiplicative gain vs. contrast gain) and to determine if feature-based attention systematically biases population response profiles in early visual cortex even before a stimulus is presented. In the Second Aim, we critically evaluate the common intuition that attention gain should be applied to sensory neurons that are maximally responsive to a target stimulus. Instead, we will test the counter-intuitive prediction that gain should sometimes be applied to neurons that are actually not tuned to the attended feature in order to maximize the efficiency of population codes. Together, these efforts will shed light on how the behavioral goals of an observer can shape population response profiles so that information processing within the visual system can be optimized. Since perception is a fundamental aspect of human information processing, our findings will be of interest to investigators focusing on perceptual learning, decision making, and memory. Moreover, the knowledge gained here will provide an important foundation for future applied research into disorders of attention such as Attention Deficit Disorder (ADD). For example, developing a better understanding about how attention modulates sensory neurons may lead to more objective diagnostic tests so that these disorders may be identified earlier and with greater accuracy. PUBLIC HEALTH RELEVANCE Whether listening to a teacher in a classroom or driving a car down the road, the ability to pay attention to important parts of the environment is critical to our success and survival. In the present research proposal, we seek to understand how patterns of neural activity in the brain support attentive behavior so that an observer may more efficiently understand and represent incoming sensory information. This knowledge will aid in the development of more objective tests for common disorders of attention - such as attention deficit disorder - so that diagnosis can proceed with greater precision and so that interventions can be started earlier.