The structure and significance of correlated activity among retinal ganglion cells

摘要 相关活动是神经元群体如何处理信息的一个组成部分。信号和噪声 考虑相关性是重要的，因为它们具有改进或降低编码的潜力， 解码，这取决于它们的具体结构。在这里，我建议利用视网膜作为一个模型系统来阐明 相关活动对视觉处理的影响。特别是，我将测试跨 不同的计算和适应状态。视网膜神经节细胞（RGC）， 大脑的视觉信息，在它们的反应中表现出信号和噪声的相关性。RGC分为 ~30种不同的细胞类型，每种都有不同的视觉功能。在这里，我考虑的结构和 不同RGC类型之间的相关性的重要性。适应状态是另一个关键因素， 视网膜必须在很宽的光强度范围内传递视觉信号，不同的光强度会改变强度。 RGC之间的相关性。我的总体假设是，RGC之间的相关活动编码了新的 视觉信号，并可以大大改善视网膜下游电路的解码。我来测试一下 假设在三个目标中跨越细胞类型和光水平。我的第一个目标是测量和模拟 RGC之间的相关性，以量化相关活动的基础反应成分。我的第二个目标， 我将确定相关活动如何影响视觉场景的视网膜编码。具体来说，我会找到 刺激特征和由跨细胞类型的相关活动编码的信息量， 光水平。第三个目标将确定相关性如何影响视网膜神经元的下游处理。 输出.在这个目标中，我将实现一个解码器，估计从RGC反应的视觉刺激。我会决定 如果相关活动提高了解码器的性能 不同光照水平下的RGC活动。总的来说，这项工作利用并结合了大规模 多电极阵列与先进的计算技术。这些方法将使我能够确定 视网膜中编码和解码视觉刺激的相关活动的作用。这项工作将促进我们的 早期视觉处理的知识，以及相关性如何影响神经计算和电路功能 梗概.

ABSTRACT Correlated activity is an integral part of how populations of neurons process information. Signal and noise correlations are important to consider because they have the potential to improve or degrade encoding and decoding, depending on their specific structure. Here I propose to utilize the retina as a model system to elucidate the impact of correlated activity on visual processing. In particular, I will test the role of correlations across different computations and adaptation states in the retina. Retinal ganglion cells (RGCs), the sole source of visual information for the brain, exhibit signal and noise correlations in their responses. RGCs are organized into ~30 different cell types, each of which performs a distinct visual function. Here I consider the structure and significance of correlations across diverse RGC types. Adaptation state is another crucial factor because the retina must convey visual signals over a broad range of light intensities, and different light levels alter the strength of correlations between RGCs. My overarching hypothesis is that correlated activity among RGCs encodes novel visual signals and can substantially improve decoding by circuits downstream of the retina. I will test this hypothesis across cell types and light levels in three aims. My first aim will measure and model the structure of correlations among RGCs to quantify response components that underlie correlated activity. In my second aim, I will determine how correlated activity impacts retinal encoding of visual scenes. Specifically, I will find the stimulus features and the amount of information that are encoded by correlated activity across cell types and light levels. The third aim will determine the how correlations may impact downstream processing of retinal output. In this aim, I will implement a decoder that estimates visual stimuli from RGC responses. I will determine if correlated activity improves the decoder’s performance such that a single decoder can successfully readout RGC activity across light levels. Overall, this work utilizes and combines the powerful capabilities of large-scale multi-electrode arrays with advanced computational techniques. These approaches will enable me to determine the role of correlated activity for encoding and decoding visual stimuli in the retina. This work will advance our knowledge of early visual processing, as well as how correlations impact neural computations and circuit function in general.