Deep sampling of cognitive effects in the human visual system

人类视觉系统认知效应的深度采样

• 批准号: 10658424
• 负责人: CLAYTON E CURTIS
• 金额: $ 28.74万
• 依托单位: UNIVERSITY OF MINNESOTA
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-07-01 至 2027-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10658424
• 关键词: Algorithms; Area; Attention; Behavior; Benchmarking; Brain; Cognitive; Complex; Computer Models; Data; Data Set; Decision Making; Development; Dyslexia; Estimation Techniques; Event; Experimental Designs; Functional Magnetic Resonance Imaging; Goals; Grant; Human; Image; Individual; Instruction; Interdisciplinary Study; Investigation; Investments; Laboratories; Libraries; Linear Models; Maps; Measurement; Mediating; Methodology; Methods; Modeling; Nature; Neural Network Simulation; Noise; Participant; Performance; Primates; Prosopagnosia; Research; Research Personnel; Resolution; Resources; Response to stimulus physiology; Sampling; Sensory; Sensory Disorders; Short-Term Memory; Signal Transduction; Software Tools; Source; Specificity; Stimulus; Support System; Techniques; Temporal Lobe; Testing; Variant; Venous; Visual; Visual Cortex; Visual Perception; Visual System; Work; blood oxygenation level dependent response; cognitive control; cognitive neuroscience; cognitive process; cognitive task; computational neuroscience; data resource; data reuse; deep neural network; experience; experimental study; extrastriate visual cortex; frontal lobe; hemodynamics; improved; information processing; interdisciplinary approach; intraparietal sulcus; multidisciplinary; neural; neuroimaging; novel; predictive modeling; programs; response; success; theories; tool; visual information; visual neuroscience; visual processing; visual stimulus

PROJECT SUMMARY Much progress has been achieved in building computational models that describe how visual stimuli are encoded and transformed across the hierarchy of visual areas in the primate brain. However, relatively little attention has been given to the active component of visual processing, whereby the observer’s engagement in a cognitive or perceptual task can exert substantial influence on stimulus-evoked activity. The long-term goal of the proposed research is to develop models that account for not only stimulus-driven but also task-driven effects in the visual system. To achieve this goal, we invest effort in improving analysis methodology and in generating large high- quality experimental datasets. In Aim 1, we develop and optimize methods that provide robust fMRI measurements at the level of single trials. This includes a method that exploits temporal dynamics to isolate BOLD signals that are more closely related to local neural activity, and an algorithm that improves signal-to- noise ratio in general linear modeling of fMRI data. In Aim 2, we acquire and prepare a high-field (7T) fMRI dataset in which a large variety of tasks are performed on a common set of visual stimuli. This dataset exploits deep sampling of a small number of subjects, and will generate a rich, reusable resource for the fields of cognitive and computational neuroscience. In Aim 3, we exploit the dataset to test an extant model of how top-down influences from the intraparietal sulcus modulate responses in ventral visual cortex and to assess how top-down signals from frontal cortex modulate the fidelity of working memory stimulus encoding in visual cortex. Overall, this multidisciplinary proposal will deliver reusable methods and data resources, and will push computational modeling from the domain of stimulus representation into better understanding how the visual system mediates real-world visual perception and task engagement. Advancing our understanding of the computational mechanisms underlying visual processing in healthy individuals is a critical step for unraveling the nature of sensory disorders such as prosopagnosia and dyslexia.

项目摘要 在建立描述视觉刺激如何编码的计算模型方面已经取得了很大进展 并在灵长类动物大脑的视觉区域中进行了转换。然而，相对较少的关注 被赋予视觉处理的主动成分，从而观察者参与认知或 知觉任务对刺激诱发活动有重要影响。拟议的长期目标 研究的目的是开发模型，不仅解释刺激驱动的，而且解释任务驱动的视觉效果。 系统为了实现这一目标，我们致力于改进分析方法，并产生大量高质量的数据， 高质量的实验数据。在目标1中，我们开发并优化了提供强大功能磁共振成像的方法 在单次试验的水平上进行测量。这包括一种利用时间动态来分离 BOLD信号与局部神经活动更密切相关，以及一种改善信号与神经活动关系的算法。 噪声比一般线性建模的fMRI数据。在目标2中，我们获得并准备了高场（7T）fMRI 数据集，其中对一组常见的视觉刺激执行各种任务。这个数据集利用了 深度抽样的少数科目，并将产生丰富的，可重复使用的资源，为领域的认知 和计算神经科学。在目标3中，我们利用数据集来测试一个现存的模型， 顶内沟的影响调节腹侧视觉皮层的反应，并评估如何自上而下 来自额叶皮层的信号调节视皮层中工作记忆刺激编码的保真度。总的来说， 这一多学科提案将提供可重复使用的方法和数据资源，并将推动计算 从刺激表征的领域建模，以更好地理解视觉系统如何介导 真实世界的视觉感知和任务参与。推进我们对计算的理解 健康个体视觉处理的潜在机制是揭示 感觉障碍，如面容失认症和阅读障碍。