The neural representation and transformation of color in human visual cortex

描述（由申请人提供）：任何刺激的表示都可以被视为大量神经元中独特的分布式活动模式。这些神经表征被认为在视觉皮层的处理阶段经历了一系列的转换，并取决于行为需求。这些转换的性质揭示了重要的见解，从传入的感觉信号的行为有意义的神经表征的形成的计算机制。在这个建议中，重点是测量神经表征和表征特定视觉模态的转换：颜色。颜色的表现形式有很多种。例如，人类区分成千上万的色调，但只使用少数离散的颜色类别。这使得颜色成为研究分布式神经表征的理想候选者。 当前提案中的新的经验和理论方法旨在显着推进对以下内容的理解：1）人类视觉系统如何表示颜色，2）这种分布式神经表示如何在视觉皮层区域的层次结构中转换，3）这些表示对行为需求的依赖性，以及4）对上下文的依赖性。目的1实验将检验颜色的神经表征随着颜色信号进入视觉系统而发生变化的假设。神经颜色空间将来自功能性磁共振成像（fMRI）测量，使用新的实验协议和多变量数据分析技术。这些神经色彩空间将与来自颜色辨别和分类的心理物理测量的感知色彩空间进行比较。目标1实验还将测试颜色的神经表征取决于行为需求的假设。所提出的实验将区分两个特定的计算假设：1）神经颜色空间针对不同的行为任务而改变，这表明神经颜色空间的潜在选择性和调谐的变化。 神经元，相对于2）响应增益的（可能选择性的）增加，没有颜色空间变化的证据。目标2实验将测试的假设，颜色感知的变化，由于一个戏剧性的视错觉，与相应的移位在底层的神经表示，和移位的程度在视觉区域之间的变化，取决于在每个视觉区域的神经颜色空间。 最终，颜色为更复杂的神经表征提供了模型（例如，那些潜在的面部和物体识别、运动控制等）这些发现将为分布式神经过程和表征提供一般性的见解。因此，拟议中的研究将提供有关大脑如何将一组传入的信号（任何模态）转换为一组有意义的表征的信息，这些表征有助于完成多种任务。 公共卫生相关性：目前的建议建立在以前的工作，研究大脑如何表示视觉信息，大脑执行的计算来转换视觉信息的表示，以及神经活动和视觉感知之间的关系。因此，所概述的方法将在理解弱视、低视力和感知性视觉失认症方面具有很大的益处，因为在该提议中开发的方法可以用于表征除了颜色之外的其他刺激特征的分布式神经表征，包括例如对比度、方向、深度、运动等。在人脑中的非侵入性，将有助于评估任何用于视网膜恢复的疗法的功效，无论其是用于弱视中的皮质视觉缺陷的疗法还是用于退行性视网膜疾病的视网膜假体。

DESCRIPTION (provided by applicant): The representation of any stimulus can be seen as a unique and distributed pattern of activity in a large population of neurons. These neural representations are thought to undergo a series of transformations across processing stages in visual cortex, and to depend on behavioral demands. The nature of these transformations reveals important insights into the computational mechanisms underlying the formation of behaviorally meaningful neural representations from incoming sensory signals. In this proposal, the focus is on measuring the neural representations and characterizing the transformations for a specific visual modality: color. The representation of color takes on many different forms. For example, humans discriminate between many thousands of hues but use only a handful of discrete color categories. This makes color an ideal candidate to investigate distributed neural representations. The novel empirical and theoretical approaches in the current proposal aim to significantly advance understanding of 1) how the human visual system represents color, 2) how this distributed neural representation is transformed across the hierarchy of visual cortical areas 3) the dependence of these representations on behavioral demands, and 4) the dependence on context. Aim 1 experiments will test the hypothesis that the neural representation of color is transformed as chromatic signals ascend the visual system. Neural color spaces will be derived from functional magnetic resonance imaging (fMRI) measurements, using novel experimental protocols and multivariate data analysis techniques. These neural color spaces will be compared with perceptual color spaces derived from psychophysical measurements of color discrimination and categorization. Aim 1 experiments will also test the hypothesis that neural representations of color depend on behavioral demands. The proposed experiments will distinguish between two specific computational hypotheses: 1) that neural color spaces change for different behavioral tasks, indicating a change in the underlying selectivity and tuning of the neurons, versus 2) a (possibly selective) increase in response gain, with no evidence for a change in the color space. Aim 2 experiments will test the hypothesize that changes in color perception, due to a dramatic visual illusion, are correlated with corresponding shifts in the underlying neural representation, and that the extent of the shift in the neural representation varies between visual areas, depending on the neural color space in each visual area. Ultimately, color provides a model for more complex neural representations (e.g., those underlying face and object recognition, control of movement, etc.) and the findings will provide general insights about distributed neural processes and representations. Consequently, the proposed research will provide information about how the brain transforms an incoming set of signals (of any modality) into a set of meaningful representations that subserve a multitude of tasks. PUBLIC HEALTH RELEVANCE: The current proposal builds on previous work to study how the brain represents visual information, the computations the brain performs to transform representations of visual information, and the relationship between neural activity and visual perception. Consequently, the outlined approach will be of great benefit in understanding amblyopia, low vision, and aperceptive visual agnosias, as the methods developed in the proposal can be used to characterize distributed neural representations of other stimulus features besides color including, for example, contrast, orientation, depth, motion, etc. In addition, these methods for characterizing large-scale, distributed neural representations, non-invasively in the human brain, will be helpful for evaluating the efficacy of any therapy for visio restoration, whether it be a therapy for cortical visual deficits in amblyopia or a retinal prostheis for degenerative retinal diseases.