Neural Visual Coding: Image to Object Representation

描述(由申请者提供)：关于视觉有一些我们不理解的基本东西。日常经验告诉我们，我们可以感知到一个物体的每一个细节，我们的眼睛可以分辨出来，即使这个物体很复杂。但神经生理学证据表明，最详细的视觉信息只出现在早期处理阶段(V1和V2区)，在那里神经元有固定在视网膜上的小感受区。这意味着每次眼睛移动到新的注视状态时，一组不同的神经元都会被激活，这种情况每秒持续发生3-4次。如何从这些分散的响应中获得对对象的稳定感知？这项申请提出了一项研究，以测试一种新的理论，该理论描述了视觉大脑中对象表征的出现及其在注意力分配中的作用。根据该理论，最近在V1和V2中发现的嘴唇和纹外视觉区域感受野的重新映射和图形-背景结构的重新映射具有共同的机制。拟议中的实验将使用单细胞录音来表现非人类灵长类动物的行为。通过记录顶叶和纹状体外的视觉区域，该理论的两个关键预测将得到验证。一种是感受野的重新映射，就像经典的眼跳观察到的那样，也应该随着物体的移动而发生。将对使用经典重新映射范例发现重新映射的几个领域进行测试，预计不同领域的结果会有所不同。第二个预测是，在高电平(例如，顶区)和低电平(例如，V2区)记录的重新映射信号应该表现出特定的相关性。这将通过用两个微电极同时记录来进行测试。如果得到证实，这将是朝着包括图形-背景组织、基于对象的注意和知觉稳定性的视觉处理综合理论迈出的一步。

DESCRIPTION (provided by applicant): There is something basic about vision that we do not understand. Everyday experience tells us that we can perceive every detail of an object that our eyes can resolve, even if the object is complex. But neurophysiological evidence shows that the most detailed visual information is represented only at the early processing stages (areas V1 and V2), where neurons have small receptive fields that are fixed on the retinae. This means that a different set of neurons is activated every time the eyes move to a new fixation, which occurs continually 3-4 times per second. How can a stable percept of the object result from these scattered responses? This application proposes research to test a new theory that describes the emergence of object representations in the visual brain and its role in the deployment of attention. According to the theory, the remapping of receptive fields as found in LIP and extrastriate visual areas, and the remapping of figure-ground structure that was recently discovered in V1 and V2, have a common mechanism. The proposed experiments will use single-cell recordings in behaving non-human primates. By recording from parietal and extrastriate visual areas, two critical predictions of the theory will be tested. One is that remapping of receptive fields, as classically observed in connection with saccades, should also occur in response to object movements. Several areas where remapping has been found with the classic remapping paradigm will be tested, and results are expected to differ between areas. The second prediction is that the remapping signals recorded at high level (e.g., parietal area) and low level (e.g., area V2) should exhibit specific correlations. This will be tested by recordin simultaneously with two microelectrodes. If confirmed, this would be a step towards a comprehensive theory of visual processing that includes figure-ground organization, object-based attention, and perceptual stability.