Neurophysiology of Visual Perception

人类完全致力于使用视觉进行社会互动。例如，人类漫长的童年时期教会我们读懂非常微妙的面部表情，这些表情可以告诉我们一个人是在撒谎、生气、尴尬、怀疑还是在匆忙。此外，视觉是人类另一项重要技能的核心：以无数方式使用我们的双手的能力。 我们不仅在社会交往中使用视觉来指导我们的手，而且我们还能够在使用工具，参与体育运动或演奏乐器方面更进一步。 这种人类的例外论建立在许多其他哺乳动物物种所共有的令人印象深刻的视觉系统之上，它们经常使用视觉来识别领土，筑巢，躲避捕食者和寻找配偶。虽然动物的这些日常问题有时被研究人员淡化为仅仅涉及先天行为，但所涉及的视觉问题往往非常复杂，与人类视觉有许多共同之处。 虽然我们已经对视觉大脑有了很多了解，但视觉的许多基本问题，以及它们对我们如何看待和解释他人以及我们周围的世界的影响，仍然知之甚少。 我们在视觉感知方面的大部分实验工作都集中在大脑如何分析社会刺激和场景上。 对于视觉来说，这一过程必然始于视网膜，其内容通过一系列高度专门化的皮层区域进行解释。 大脑不仅利用这些信息来识别物体，而且还为感知和行动创造了一个三维的内部世界。它流畅地执行这一操作，保持稳定的视觉场景，同时以某种方式抵抗由我们自己的运动引起的持续视网膜干扰，最明显的是眼睛注视的变化。 在一个主要项目中，我们一直在研究猕猴颞叶皮层中的神经元在自由运行范式中的反应，这种范式与传统的测试模式不同，传统的测试模式是短暂闪现的图像刺激的连续呈现。 在这个测试中，我们允许受试者观看自然视频，每次播放几分钟。受试者可以自由地浏览视频内容，我们会记录他们的目光和高级视觉皮层不同区域的许多神经元的活动。我们的分析也与传统不同，因为我们不关注刺激表征本身，而是关注大脑的不同部分如何在处理不同类型的场景时协同工作。为此，我们将我们的多个单一单元记录与功能性MRI（fMRI）的数据相结合，其中猕猴受试者观看相同的自然视频。由于功能性磁共振成像能够覆盖全脑，我们就能够使用单个单位活动作为一种种子或回归因子来获得功能性磁共振成像数据的全脑地图。发表的第一篇使用这种方法的论文强调了颞叶皮层神经元局部群体内的局部反应多样性。也许最令人惊讶的是，它表明，当通过上述种子相关方法进行分析时，相邻的神经元参与了非常不同的全脑网络。关于这个主题的第二份手稿，比较多个颞叶皮层区域的当地人口，目前正在编写中。 在另一个项目中，我们研究了自然场景流动过程中时间结构的性质。 具体来说，我们问的问题是否时间动态本身是重要的决定在神经的选择性。为此，我们从连续的电影中提取了一秒钟的片段，并将提取的组件的神经反应与电影完整显示时产生的神经反应进行了比较。我们的研究结果表明，从一个时刻到另一个时刻的时间整合是许多神经元放电的重要决定因素。此外，最初的视觉反应瞬变显示出与神经元对完整电影的相同时刻的反应不相关的刺激调谐。后一个发现非常令人惊讶，可能会对我们如何看待视觉大脑产生深远的影响，因为我们对皮层微电路、视觉层次、刺激选择性和功能结构的大多数理解都来自于刺激在屏幕上短暂闪现的实验。 在我们的枕的研究中，我们继续调查的空间分布的神经元响应特定的视觉功能在这个大的核复合体。在一项研究中，我们测量了皮质顶盖束及其周围的神经元浓度，这些神经元对面孔有明显的选择性。在目前正在准备的一份手稿中，我们将这些面部选择性反应的性质与颞叶皮层的fMRI面部斑块中所谓的面部细胞进行了比较。主要发现之一是，许多枕面部选择性神经元的反应比面部斑块中的最早反应早得多，这表明它们的面部选择性不能简单地从这些区域遗传。 在另一项研究中，我们测量了枕神经元对自然视频的反应，如上文所述的面部斑块。 我们发现一些枕神经元表现出类似的时间进程和fMRI映射配置文件的脸补丁神经元。 这一观察结果，连同他们的非常短的反应lavelet，提出了重要的问题枕在处理某些，专门的刺激，如面孔的作用。 一种可能性是枕神经元有两个重要的作用：首先，它们通过次级视觉通路将这些视觉信号传递到皮层，其次，它们接收足够的皮层输入，以协调快速枕通路和较慢的膝状皮质通路。

Humans are fully committed to the use of vision for their social interaction. For example, the extended childhood of humans teaches us to read very subtle facial expressions that can tell us whether a person is lying, angry, embarrassed, skeptical, or in a hurry. Further, vision is at the center of another important human skill: the capacity to use our hands in countless ways. Not only do we use vision to direct our hands during social interaction, but we are also able take this much further in the use of tools, participation in sports, or the playing musical instruments. This human exceptionalism rests atop an impressive visual repertoire shared by many other mammalian species, who use often use vision to recognize territories, build nests, avoid predators, and find mates. While these everyday problems of animals are sometimes downplayed by researchers as simply involving innate behaviors, the visual problems involved are often highly complex and share much in common with human vision. While much has been learned about the visual brain, many of the basic problems of vision, and their bearing on how we see and interpret others and the world around us, remain poorly understood. Much of our experimental work on visual perception centers on how the brain analyzes social stimuli and scenes. For vision, this process necessarily begins in the retina, whose contents are interpreted though a series of highly specialized cortical areas. The brain uses this information not only to recognize objects, but also to create a three-dimensional, internal representation of the world for perception and action. It performs this operation fluidly, maintaining a stable visual scene while somehow resisting distraction by the continuous retinal disturbances caused by our own movements, most notably changes in eye gaze. In one major project, we have been investigating how neurons in the macaque temporal cortex respond during free-running paradigm that depart from the conventional mode of testing, which is the serial presentation of briefly flashed image stimuli. During this testing, we allow subjects to view natural videos playing out over several minutes at a time. Subjects are free to scan the content of the videos, as we record their gaze and the activity of many neurons in different regions of the high-level visual cortex. Our analysis also departs from convention in that we do not focus on stimulus representation per se, but rather on how different parts of the brain work together in processing different types of scenes. For this, we combined our multiple single-unit recordings with data from functional MRI (fMRI), in which macaque subjects watched the same natural videos. Given the full-brain coverage afforded by fMRI, we were then able obtain whole-brain maps of fMRI data using single-unit activity as a sort of seed or regressor. The first paper using this method was published emphasized the local response diversity within a local population of temporal cortex neurons. Perhaps most surprisingly, it demonstrated that neighboring neurons participated in very different whole-brain networks when analyzed through the seed correlation method mentioned above. A second manuscript on this topic, comparing local populations in multiple temporal cortex areas, is currently under preparation. In another project, we have investigated the nature of the temporal structure during the flow of a natural scene. Specifically, we have asked the question whether the temporal dynamics are themselves important for determining in neural selectivity. To this end, we extracted one-second segments from a continuous movie and compared the neural responses to the extracted components to those arising when the movie was shown intact. Our results indicate that temporal integration from moment to moment is an important determinant in the firing of many neurons. Further, the initial visual response transient showed a stimulus tuning that was uncorrelated with the neurons response to the same moments of the intact movie. This latter finding was very surprising and may have profound consequences for how we think about the visual brain, since most of our understanding of the cortical microcircuit, visual hierarchy, stimulus selectivity, and functional architecture are derived from experiments in which stimuli were flashed briefly onto a screen. In our studies of the pulvinar, we have continued to investigate the spatial distribution of neurons responsive to particular visual features across this large nuclear complex. In one study, we measured a concentration of neurons in and around the corticotectal tract that are notably selective for faces. In a manuscript currently under preparation, we have compared the nature of these face-selective responses with so-called face-cells in the fMRI mapped face patches of the temporal cortex. One of the principal findings is that many of the pulvinar face-selective neurons respond much earlier than the earliest responses in the face patches, suggesting that their face selectivity cannot be simply inherited from those areas. In another study, we measured the responses of pulvinar neurons to natural videos, as described above for face patches. We found a number of pulvinar neurons that exhibit similar time courses and fMRI-mapping profiles to the face patch neurons. This observation, together with their very short response latencies, raises important questions about the role of the pulvinar in the processing of certain, specialized stimuli such as faces. One possibility is that pulvinar neurons have two important roles: first, they pass such visual signals up to the cortex through a secondary visual pathway, and second they receive ample cortical input to coordinate the fast pulvinar pathway with the slower geniculocortical pathway.