A quantitative method of computer analysis of spike train data collected from behaving animals
A quantitative method of computer analysis of spike train data collected from behaving animals
复制标题
对从行为动物收集的尖峰序列数据进行计算机分析的定量方法
DOI：
10.1016/0006-8993(79)90530-4
复制
发表时间：
1979
期刊：
影响因子：
2.9
通讯作者：
J. Aldridge
中科院分区：
文献类型：
--
作者：
J. Macpherson;J. Aldridge
185 discrete point process to a continuous function expressing the actual density of spikes in time. The spike train can in fact be manipulated to produce a continuous function as has been briefly described by Levick and Zacks3: each spike is replaced by a curve whose maximum is at the time of occurrence of the spike and which decreases smoothly along the time axis on either side of the spike. We chose the Gaussian function to replace each spike since its properties are well-described and it is relatively simple to work with. Others have used the same sort of approach, but with hard-wired techniques. Frost and Low 1 used an integrating circuit which converted a spike pulse to an asymmetric step function with an exponential decay. The Gaussian function, a symmetrical function, requires two parameters, the mean (set at the spike time) and the width, expressed as the standard deviation (in msec). The curve representing each spike is then summed with all the others at each point in time to produce a continuous function as in Fig. 1A. The magnitude of this function reflects the spike density (spikes/sec) at each point in the time course of the single trial. Using a wider Gaussian curve produces a smoother spike density function; however, in the extreme case this function becomes too spread out, with a loss of resolution in time. The optimum width produces a balance between the function being too irregular, resembling the actual spike train itself, and being too smoothed to adequately represent local spike densities. As a simple approach to this problem, we have made the choice of width for a single trial loosely dependent on the mean interspike interval of the background period. For ease of computation we limit the choice to 4 different widths. The larger the mean interval is, the larger the width of the Gaussian employed.Having generated one continuous function to represent the spike train of a single trial we then proceed to determine excitations and/or inhibitions. Each trial has two parts, a pre-event or background control period, and a post-event period. The background period is used to calculate a mean value of the function, and the 10~ o and 90~ o points of the distribution of background values. These points are taken as thresholds for significant departures from background spike density. The post-event time period is then examined for values which exceed the 90~ level (excitations), and those which fall below the 10~ o level (inhibitions). Fig. 1A shows a trial analyzed in this manner with the threshold lines drawn in. The onset and offset of each excitation and inhibition is indicated by the arrows.
