Mining Large-Scale Neural Ensemble Recordings

要了解神经元是如何协调行动的，需要观察大型神经元在空间上的集体活动 分布的神经元聚集体。在很大程度上受到微制造技术快速进步的推动， 高密度可植入电子接口现在能够获得大量的 生理和行为数据，引发伴随而来的神经生物学发现。不过， 高密度微电极阵列(MEA)的制备进展与量子无关 为了揭示丰富的信息内容，阵列处理和数据分析技术的进展 在记录的神经数据中。随着单个微探针设备上的记录通道的数量变为 令人震惊的是，没有哪门学科比信号处理和数据挖掘更具挑战性 在新兴的神经工程领域内适应这些新的进步。有一种内在的 需要设计新的算法和软件工具来优化阵列处理和信息检索 多个棘波训练神经数据以回答几个持续存在的神经科学问题。 本研究的基本目标是探索和开发一种集成的阵列处理和 数据挖掘框架和配套的软件工具，用于从大规模数据中提取有用的信息 通过以下目标进行神经元合奏录音： 1.开发可扩展、自适应的高密度微电极阵列处理算法 短期和长期实验装置中的阵列记录； 2.开发数据分析和聚类技术，挖掘神经功能间的相互依赖关系 来自录制的混音的合奏； 3.开发了一个集成了所开发的阵列处理算法的开源软件包 和在目标2下开发的数据聚类算法，并将该包分发给 社区； 4.测试和演示这些技术的效率和开发的软件的完整性 现场研究人员分享的模拟和实验数据。 在完成拟议的研究活动后，我们预计将为众多用户提供 神经科学界拥有新的工具，可以更准确地处理和分析他们的数据， 在他们的行为实验中最大化的效率和持续的可靠性。

Understanding how neurons act in concert requires observation of the collective activity of large, spatially distributed neuronal aggregates. Largely motivated by the rapid advances in microfabrication technology, high-density implantable electronic interfaces are now enabling the acquisition of large volumes of physiological and behavioral data, triggering concomitant neurobiological discoveries. Nevertheless, advances in the fabrication of high-density microelectrode arrays (MEAs) were not associated with quantum advances in array processing and data analysis techniques in order to unveil the affluent information content in the recorded neural data. As the number of recording channels on a single microprobe device becomes astoundingly large, no discipline is more challenged than signal processing and data mining in accommodating these new advances within the emerging neural engineering arena. There is an intrinsic need to design new algorithms and software tools to optimize array processing and information retrieval from multiple spike train neural data to answer several persistent neuroscience questions. The fundamental objective of this research is to explore and develop an integrated array processing and data mining framework with companion software tools to extract the useful information from large-scale neuronal ensemble recordings through the following aims: 1. Develop scalable and adaptive array processing algorithms for processing high-density microelectrode array recordings in short and long-term experimental setups; 2. Develop data analysis and clustering techniques for mining functional interdependency among neural ensembles from the recorded mixtures; 3. Develop an open source software package that integrates the array processing algorithms developed under aim 1 with the data clustering algorithms developed under aim 2 and disseminate the package to the community; 4. Test and demonstrate the efficiency of these techniques and the integrity of the developed software on simulated and experimental data shared by investigators in the field. Upon completion of the proposed research activity, we anticipate to provide numerous users in the neuroscience community with novel tools for processing and analyzing their data with increased accuracy, maximized efficiency and sustained reliability in their behavioral experiments.