A Fast, Accurate and Cloud-based Data Processing Pipeline for High-Density, High-Site-Count Electrophysiology

在过去的十年里，神经科学家可用的工具取得了重大进展，这使得提出以下问题成为可能 越来越具体的问题，关于哪些神经元和电路与之相关，对哪些神经元和电路是必要的，以及 足以满足特定的行为或计算功能。我们对神经学的理解的进展 计算及其如何导致复杂行为在很大程度上取决于协调的 行为动物的神经活动。在过去的几年里，我们做出了重大的协调努力， 通过增加站点密度、扩大空间覆盖范围、提供高保真来推进神经探头技术 记录，并与细胞类型的特定刺激工具集成。细胞外录音显示出行动 需要棘波分类分析才能正确检测并将其分配给单个神经元的电位(棘波)。 由于细胞外的广泛可访问性，对准确和可扩展的尖峰信号分类的需求已经增加 记录技术以及记录和分离尽可能多的神经元活动的需求增加 有可能。神经科学的基本发现，如定向选择细胞、放置细胞和网格细胞 如果没有对细胞外信号进行可靠的尖峰分类，这是不可能的。这些发现已经 阐明了信息处理和认知能力的细胞基础。最近，细胞外记录 设备也被应用于通过假肢恢复运动功能。然而，由于 这些治疗所需的电极增加了，以实现更精细的电机控制，因此对自动化的需求也增加了 分析。有了更多的自动尖峰排序能力，这些领域的领域进展将是 加速了。现有的尖峰分析解决方案缺乏可扩展性，通常被设计为锁定 用户添加到特定的硬件平台。社区对集成的开源分析平台的需求是 随着细胞外电极容量的增加以及新的和未验证的电极数量的增加而快速增长 钉子排序法。JRCLUST是我们的免费、开源、独立的尖峰分类软件，它提供了一个可扩展的、 自动化且经过充分验证的尖峰排序工作流，可容忍实验记录条件 噪声、探头漂移和动物行为产生的运动伪影。它可以使用一个集合来处理各种数据集 预优化的参数，使其在社区中得到广泛应用。此外，我们的处理速度和 模块化方法允许快速循环创新和实用途径，以解释来自 数以百计的录音地点。由于其实时性能和准确的自动化分析，只需 JRCLUST是一台单一的工作站，自成立以来不到一年的时间就在全球20个实验室迅速采用 一年前。该项目的成功完成将使Vidrio能够支持、扩展和维护JRCLUST 从而使研究人员能够阐明功能定义的神经元亚群如何介导特定的 在行为过程中的关键时刻，在健康动物和在动物模型中，信息处理功能 神经系统疾病。

The past decade has seen major advances in the tools available to neuroscientists, making it possible to ask increasingly specific questions regarding which neurons and circuits are correlated with, necessary for, and sufficient for, specific behavioral or computational functions. Advancements in our understanding of neural computation and how it leads to complex behavior critically depend on accurate measurements of coordinated neural activities in behaving animals. In the past several years there have been major coordinated efforts to advance neural probe technology by increasing site density, extending spatial coverage, providing high fidelity recording, and integrating with cell-type specific stimulation tools. Extracellular recordings exhibit action potentials (spikes) that require spike-sorting analysis to correctly detect and assign them to individual neurons. The demand for accurate and scalable spike sorting has increased due to the wide accessibility of extracellular recording technology and the increased requirements to record and separate activity from as many neurons as possible. Fundamental discoveries in neuroscience such as orientation-selective cells, place cells, and grid cells would not have been possible without reliable spike sorting of extracellular signals. These discoveries have illuminated the cellular basis of information processing and cognitive abilities. Recently, extracellular recording devices have also been applied to restoring motor function through prosthetics. However, as the number of electrodes needed for these treatments increases to allow for finer motor control so does the need for automated analysis. With more capacity for automated spike sorting the field’s progress in these domains would be accelerated. Existing spike analysis solutions suffer from a lack of scalability and are often designed to lock a user into a specific hardware platform. The community’s need for an integrated open-source analysis platform is rapidly growing with the increasing capacity of extracellular electrodes and the number of new and un-validated spike-sorting methods. JRCLUST, our free, open-source, standalone spike sorting software, offers a scalable, automated and well-validated spike sorting workflow that can tolerate experimental recording conditions with noise, probe drift, and motion artifacts from behaving animals. It can handle a wide range of datasets using a set of pre-optimized parameters making it practical for wide use in the community. Also, our processing speed and modular approach allows for rapid cycle innovation and practical pathways to interpret long recordings from hundreds of recording sites. Thanks to its real-time performance and accurate automated analysis requiring only a single workstation, JRCLUST has been rapidly adopted in 20+ labs worldwide since its inception less than a year ago. Successful completion of this project will enable Vidrio to support, expand and maintain JRCLUST thus empowering researchers to elucidate how functionally defined subpopulations of neurons mediate specific information-processing functions at key moments during behavior, in healthy animals and in animal models of neurological diseases.