Prediction and Surprise in the Retina.

视网膜中的预测和惊喜。

• 批准号: 9319045
• 负责人: Lane Thomas McIntosh
• 金额: $ 4.1万
• 依托单位: STANFORD UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-09-15 至 2019-09-14
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9319045
• 关键词: Amacrine Cells; Animals; Behavior; Benchmarking; Biological; Biophysics; Cells; Code; Computer Simulation; Data; Development; Environment; Exhibits; Future; Gaussian model; Goals; Higher Order Chromatin Structure; Human; Influentials; Injectable; Injection of therapeutic agent; Interneurons; Knowledge; Linear Models; Measurement; Measures; Mission; Modeling; Neurons; Noise; Non-linear Models; Pathway interactions; Poaceae; Process; Property; Public Health; Research; Retina; Retinal Ganglion Cells; Scheme; Signal Transduction; Stimulus; Structure; System; Techniques; Testing; United States National Institutes of Health; Visual; Visual system structure; Work; design; disability; experimental study; ganglion cell; horizontal cell; improved; inhibitory neuron; innovation; movie; neural circuit; next generation; public health relevance; receptive field; response; retinal prosthesis; sensory system; spatiotemporal; success; theories; visual coding; visual stimulus

 DESCRIPTION (provided by applicant): There is a fundamental gap in understanding how the retina responds under natural conditions. Over the last few decades, the discovery of nonlinear and long-range processes in the retina have made this gap particularly troublesome, since single-pathway linear models can not adequately explain the retina's behavior under natural stimuli. The long-term goal is to situate the retina within a theoretical framework that can explai and predict its responses to natural scenes. The overall objective in this application is to extend and test one theory, predictive coding, to provide a better explanation of the retina's behavior under natural scenes. The central hypothesis is that the retina implements predictive coding via inhibitory circuits that function to remove predictable - and hence redundant - features of the natural world. This hypothesis is substantiated by evidence that the retina appears to perform predictive coding under simple Gaussian stimuli as well as preliminary evidence from intracellular current injection experiments that show correlations of amacrine- ganglion cell pairs changing in a way consistent with predictive coding theory. The hypothesis will be tested with two specific aims that 1) extend the theory of predictive coding to natural scenes, and 2) make experimental tests of predictive coding by directly recording from and injecting current into inhibitory cells in the retina. Under the first aim, I will find nonlinear, biophysically feasible models of natural scenes that minimize prediction error; from the perspective of predictive coding, these models would form the ideal inhibitory circuit. I will then compare these models to the response properties of retinal ganglion cells and amacrine cells. The second aim makes use of intracellular current injection experiments to directly measure the prediction error between inhibitory cells and their downstream excitatory targets. This proposal is innovative, in the applicant's opinion, because it leverages both theoretical and experimental techniques to improve our framework for understanding the retina's behavior. Regardless of if the test of predictive coding is positive or negative, this study will generate new questions and resolve an ongoing controversy in the field. Additionally, this proposal will contribute to the understanding of retinal ganglion responses in natural environments that are essential to the development of next-generation retinal prosthetics.

 描述(由申请人提供)：在理解视网膜在自然条件下的反应方面存在着根本的差距。在过去的几十年里，视网膜中非线性和远程过程的发现使这个缺口变得特别麻烦，因为单路径线性模型不能充分解释视网膜在自然刺激下的行为。长期目标是将视网膜定位在一个能够解释和预测其对自然场景的反应的理论框架内。这个应用程序的总体目标是扩展 并测试预测编码理论，以更好地解释视网膜在自然场景下的行为。中心假设是，视网膜通过抑制电路实现预测性编码，该电路的功能是移除自然界的可预测特征，从而消除冗余特征。这一假说得到以下证据的证实：视网膜似乎在简单的高斯刺激下执行预测性编码，以及来自细胞内电流注入实验的初步证据，这些证据表明无长突神经节细胞对之间存在关联 以符合预测编码理论的方式改变。这一假设将以两个具体目标进行测试：1)将预测编码理论扩展到自然场景；2)通过直接从视网膜抑制细胞记录电流并向其注入电流，对预测编码进行实验测试。在第一个目标下，我将找到自然场景的非线性、生物物理上可行的模型，将预测误差降至最低；从预测编码的角度来看，这些模型将形成理想的抑制电路。然后，我将把这些模型与视网膜神经节细胞和无长突细胞的反应特性进行比较。第二个目的是利用细胞内电流注入实验，直接测量抑制性细胞与其下游兴奋靶点之间的预测误差。在申请人看来，这项建议是创新的，因为它利用了理论和实验技术来改进我们理解视网膜行为的框架。无论预测编码测试是积极的还是消极的，这项研究都将产生新的问题，并解决该领域持续存在的争议。此外，这项提议将有助于理解自然环境中视网膜神经节的反应，这对开发下一代视网膜假体至关重要。