Neural Visual Coding: Image to Object Representation

描述（由申请人提供）：本研究旨在加深我们对视觉知觉组织的神经机制及其与选择性注意的关系的理解。长期目标是在神经信号水平上理解感觉表征和认知过程之间的接口。提出的研究重点是图-地组织的机制，从边界所有权的神经表示中可以看出，即对物体的轮廓分配。一般的方法是从清醒行为的猕猴的视觉皮层进行单细胞记录。已经确定了未来五年的三大目标。第一个目的是提供边界所有权选择V2神经元的空间整合机制的定量描述。先前的研究表明，这些机制远远超出了神经元的经典接受野。将采用两种析因设计方法。这些方法允许测量非经典区域与经典接受野的非线性相互作用以及非经典区域之间的相互作用。这两种方法是互补的。一种是使用人物的碎片化，另一种是使用关键特征的遮挡。第二个目的是研究图-地组织在视觉皮层中的持久性。实验旨在(1)确定图形-背景线索的类型及其持续时间对边界所有权信号随后持续的影响，(2)确定这种持续是否取决于对图形的注意力，以及(3)边界所有权信号是否在眼球运动和刺激运动中持续存在。第三个目的是研究神经组织机制在基于对象的选择性注意中的作用。先前的结果显示，V2神经元将边界所有权编码与意志选择注意的影响结合起来，在此基础上，新的实验将测试视觉条件，在这种条件下，一个物体被部分遮挡，因此只有两个不相连的部分仍然可见。结果将显示边界所有权机制是否可以连接跨越差距的特征，以及基于对象的注意力是否可以做到这一点。本研究将有助于澄清选择性注意的基础，这对理解正常视力和受损视力都很重要。虽然该建议主要集中在视觉上，但该研究将有助于识别其他形式的感觉-认知界面。从这个项目中获得的见解将有助于理解认知障碍如阅读障碍的基础。选择性注意可能是视力中最重要的功能。我们可以选择处理一行文本中的单词，而忽略所有其他的，这使我们能够阅读。但是有时我们看不到眼前的东西，这就是许多交通事故的原因。有诵读困难的孩子无法分辨出其他人很容易看到的字母串之间的区别。这项拟议研究的目标是了解大脑中在物体被识别之前将图像分解成物体的过程，为注意力提供一个可以选择的内部表征。

DESCRIPTION (provided by applicant): Research is proposed to advance our understanding of the neural mechanisms of visual perceptual organization and their relation to selective attention. The long-term objective is to understand the interface between sensory representations and cognitive processes at the neural signal level. The proposed research focuses on the mechanisms of figure-ground organization as evident from the neural representation of border ownership, i.e., the assignment of contours to objects. The general method is single-cell recording from the visual cortex of awake behaving macaques. Three broad aims have been identified for the next five years. The first aim is to provide quantitative descriptions of the spatial integration mechanisms of border ownership selective V2 neurons. Previous studies have shown that these mechanisms extend far beyond the classical receptive field of the neurons. Two methods with factorial design will be applied. These methods allow to measure nonlinear interactions of the extra-classical regions with the classical receptive field as well as interactions between extra-classical regions. The two methods are complementary. One uses fragmentation of figures, the other uses occlusion of critical features. The second aim is to study persistence of figure-ground organization in the visual cortex. Experiments are proposed to (1) determine the influence of the type of figure-ground cue and its duration on the subsequent persistence of border ownership signals, (2) determine whether the persistence depends on attention being directed to the figure, and (3) whether border ownership signals persist across eye movements and stimulus movements. The third aim is to study the role of the neural organization mechanisms in object-based selective attention. Building on previous results showing that V2 neurons combine border ownership coding with the influence of volitional selective attention, new experiments will test visual conditions in which an object is partially occluded, so that only two unconnected portions of it remain visible. The results will show whether the border ownership mechanisms can link features across the gap, and whether object-based attention can do so. The proposed research will help to clarify the basis of selective attention which is important for understanding both normal and impaired vision. Although the proposal is focused on vision, the research will help to identify the sensory-cognitive interface also in other modalities. The insight gained from this project will contribute to the understanding of the basis of cognitive disorders such as dyslexia. PUBLIC HEALTH RELEVANCE Selective attention is perhaps the most important function in vision. We can choose to process the words in one line of text and ignore all others, which enables us to read. But sometimes we do not see things that are right in front of our eyes-the cause of many traffic accidents. Dyslexic children cannot detect the difference between strings of letters that others can easily see. The goal of the proposed research is to understand a process in the brain that dissects images into objects before objects are recognized, providing an internal representation from which attention can select.