Department of Neurobiology and Anatomy, University of Texas-Houston Medical School, Houston, Texas, USA.
J Neurophysiol. 2010 May;103(5):2912-30. doi: 10.1152/jn.00518.2009. Epub 2009 Dec 23.
Experimental advances allowing for the simultaneous recording of activity at multiple sites have significantly increased our understanding of the spatiotemporal patterns in neural activity. The impact of such patterns on neural coding is a fundamental question in neuroscience. The simulation of spike trains with predetermined activity patterns is therefore an important ingredient in the study of potential neural codes. Such artificially generated spike trains could also be used to manipulate cortical neurons in vitro and in vivo. Here, we propose a method to generate spike trains with given mean firing rates and cross-correlations. To capture this statistical structure we generate a point process by thresholding a stochastic process that is continuous in space and discrete in time. This stochastic process is obtained by filtering Gaussian noise through a multivariate autoregressive (AR) model. The parameters of the AR model are obtained by a nonlinear transformation of the point-process correlations to the continuous-process correlations. The proposed method is very efficient and allows for the simulation of large neural populations. It can be optimized to the structure of spatiotemporal correlations and generalized to nonstationary processes and spatiotemporal patterns of local field potentials and spike trains.
实验进展使得能够同时记录多个位置的活动，这大大提高了我们对神经活动时空模式的理解。这种模式对神经编码的影响是神经科学中的一个基本问题。因此，具有预定活动模式的尖峰列车的模拟是研究潜在神经编码的重要组成部分。这种人为产生的尖峰列车也可以用于在体外和体内操纵皮质神经元。在这里，我们提出了一种生成具有给定平均发放率和互相关的尖峰列车的方法。为了捕获这种统计结构，我们通过将连续空间和离散时间的随机过程阈值化来生成一个点过程。通过通过多变量自回归 (AR) 模型过滤高斯噪声来获得这个随机过程。AR 模型的参数是通过将点过程相关性转换为连续过程相关性来获得的。所提出的方法非常高效，允许模拟大型神经元群体。它可以针对时空相关性的结构进行优化，并推广到非平稳过程以及局部场电位和尖峰列车的时空模式。
