Methodological Research and Statisitical Modelling in Neuroscience

神经科学的方法论研究和统计模型

• 批准号: RGPIN-2018-04376
• 负责人: Ramezan, Reza
• 金额: $ 1.53万
• 依托单位: University of Waterloo
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=678477
• 关键词: Methodological; Research; Statisitical; Modelling; Neuroscience

The brain is, arguably, the most sophisticated organ in the human body. With approximately 100 billion nerve cells (a.k.a. neurons), each potentially connected (through synapses) to thousands of other neurons, the brain is a complex network of interconnected cells. These cells communicate through sequences of electrochemical waves, which are called spike trains. This research proposal focuses on multivariate, multiscale, and Bayesian extensions of state-of-the-art models in statistical modelling of neural spike trains. Recent data acquisition developments allow experimentalists to simultaneously record the electrical activity of several hundreds of neurons. This highlights the need for efficient data visualization methods, and multivariate statistical theories tailored for neural spike trains. Following my main research interests, i.e. computational neuroscience, this proposal focuses on four areas unified by the ultimate goal of understanding information coding processes in the brain. In particular, 1) we will develop a multivariate point process framework (multivariate Skellam Process with Resetting) for the analysis of multiple neural spike trains, which is both biologically justifiable and computationally efficient. 2) Within the multivariate point process framework developed in (1), we will develop models to capture over and under dispersed spiking activity (relative to Poisson) observed in real data. Since under-dispersed models for neural spike trains are rare and not well-studied in the literature, the outlined research will be impactful to the discipline. 3) We will develop multiscale multivariate models to capture neural spiking patterns from multiple time scales. Such patterns are frequently observed (simultaneously) in real data across different brain regions, and developing statistical models to study such data are necessary. 4) We will develop Bayesian extensions of the methods proposed in the previous three research activities. The above-mentioned research activities will build on and extend my previous work on the analysis of neural spike trains, and provide excellent opportunities for undergraduate, masters, and doctoral level training in this cutting-edge multidisciplinary research area.

大脑可以说是人体最复杂的器官。大脑拥有大约 1000 亿个神经细胞（又称神经元），每个神经细胞都可能（通过突触）连接到数千个其他神经元，因此大脑是一个由相互连接的细胞组成的复杂网络。这些细胞通过电化学波序列进行通信，这些电化学波称为尖峰序列。本研究提案重点关注神经尖峰序列统计建模中最先进模型的多变量、多尺度和贝叶斯扩展。最近的数据采集进展使实验人员能够同时记录数百个神经元的电活动。这凸显了对高效数据可视化方法以及针对神经脉冲序列量身定制的多元统计理论的需求。根据我的主要研究兴趣，即计算神经科学，该提案重点关注四个领域，这些领域的最终目标是理解大脑中的信息编码过程。特别是，1）我们将开发一个多元点处理框架（带有重置的多元Skellam Process）来分析多个神经尖峰序列，这在生物学上是合理的，并且在计算上是高效的。 2）在（1）中开发的多元点过程框架内，我们将开发模型来捕获在实际数据中观察到的分散尖峰活动（相对于泊松）。由于神经尖峰序列的分散不足模型很少见，并且在文献中也没有得到充分研究，因此概述的研究将对该学科产生影响。 3）我们将开发多尺度多元模型来捕获多个时间尺度的神经尖峰模式。这种模式经常在不同大脑区域的真实数据中（同时）观察到，因此有必要开发统计模型来研究这些数据。 4）我们将对前三项研究活动中提出的方法进行贝叶斯扩展。上述研究活动将建立并扩展我之前在神经尖峰序列分析方面的工作，并为这一前沿多学科研究领域的本科生、硕士和博士水平的培训提供极好的机会。