Visual-motor integration in smooth pursuit eye movements

摘要 许多行为在贝叶斯框架中运行，其中行动由以下因素之间的复杂交互指导： 当前的感觉信息和过去的经验，或“先验”。当当前感觉信息较弱时， 有利的是允许引导行为之前。然而，先前经验的价值随着感官的减少而减少。 信息增强。我们的目标是确定贝叶斯行为是如何从神经网络的操作中产生的。 电路.平稳追踪眼球运动是一个相对简单的感觉运动行为的例子， 像贝叶斯一样。平滑追踪是一种视觉引导下的自主眼球运动， 分为两个主要部分：视觉驱动和视觉增益。把这两个组成部分 从贝叶斯的角度来看：视觉驱动来自中颞视觉区（MT）， 从感觉数据导出的似然函数;视觉增益由平滑追踪区域控制 前视野（FEFsem），并表示先验。这两个概念及其神经实例是 集成在追踪系统中，以驱动最终的眼球运动。这两种成分都进行了研究 但我们的目标是研究它们的整合。脑桥背外侧核（DLPN）和 脑桥被盖网状肌（NRTP）都接受来自皮层追踪区的会聚输入， 一系列的视觉和视觉信号。因此，我们将记录DLPN中的单个神经元， 用清醒行为恒河猴的NRTP研究追踪过程中视觉和视觉信号的整合 系统为了更好地理解视觉和视觉信号的整合与 随着贝叶斯行为的出现，我们已经发展出一种行为范式，使我们能够快速适应 目标速度的先验信息我们可以通过调整目标的统计量来严格控制先验的适应性 速度，我们可以通过调节视觉运动信号的强度来控制先验的表达。这 将揭示视觉和视觉信号在脑桥中的整合是如何随着脑功能状态的变化而变化的。 感知信息的先验和强度。为了直接研究FEFsem在整合这些 信号，我们将我们的脑桥记录与FEFsem中多个单个单元的同时记录配对。 脑桥和FEFsem神经元的反应之间的神经元-神经元相关性将揭示功能 区域之间的连通性，并转向描述追捕电路如何作为一个系统工作。通过 为了更好地理解追踪系统如何整合先验信息和感官信息，我们将 制定感觉运动大脑回路的一般原则。

Abstract Many behaviors operate in a Bayesian framework, where actions are guided by a complex interaction between current sensory information and past experience, or “priors”. When current sensory information is weak, it is advantageous to allow the prior to guide behavior. However, the value of prior experience lessens as sensory information strengthens. Our goal is to determine how Bayesian behavior arises from the operation of a neural circuit. Smooth pursuit eye movement is an example of a relatively simple sensory-motor behavior that operates in a Bayesian-like manner. Smooth pursuit is a visually-guided voluntary eye movement that can be separated into two primary components: visual drive and visuomotor gain. It is appealing to think of these two components from a Bayesian perspective: the visual drive arises from the middle temporal visual area (MT) and represents the likelihood function derived from sensory data; the visuomotor gain is controlled by the smooth pursuit region of the frontal eye field (FEFsem) and represents the prior. The two concepts and their neural instantiations are integrated within the pursuit system to drive the ultimate eye movement. Both components have been studied independently, but our goal is to study their integration. The dorsolateral pontine nucleus (DLPN) and the nucleus reticularis tegmenti pontis (NRTP) both receive convergent input from cortical areas of pursuit and have cells with a range of visual and visuomotor signals. Therefore, we will record from single neurons in the DLPN and NRTP of awake behaving rhesus monkeys to study the integration of visual and visuomotor signals in the pursuit system. To better understand the relationship between the integration of visual and visuomotor signals and the emergence of Bayesian-like behavior, we have developed a behavioral paradigm that allows us to rapidly adapt priors for target speed. We can control tightly the adaptation of the prior by adjusting the statistics of the target speed, and we can control the expression of the prior by adjusting the strength of the visual motion signals. This will reveal how the integration of visual and visuomotor signals in the pons changes as a function of the state of the prior and the strength of sensory information. To directly study the role of FEFsem in the integration of these signals, we will pair our pontine recordings with simultaneous recordings of multiple single units in the FEFsem. Neuron-neuron correlations between the responses of pontine and FEFsem neurons will reveal functional connectivity between the areas and move towards a description of how the pursuit circuit works as a system. By gaining a better understanding of how the pursuit system integrates priors and sensory information, we will develop general principles of sensory-motor brain circuits.