锋电位功能网络构建与鸽子转向行为解码

发布时间：2018-02-23 22:03

本文关键词： 锋电位神经元功能网络归一化互信息度量指标解码　出处：《郑州大学》2017年硕士论文　论文类型：学位论文

【摘要】：脑网络是自然界最复杂的网络之一。大脑中数以亿计的神经元构成了庞大而复杂的脑结构网络,结构网络上不同神经元、神经元集群或脑区之间动态活动的同步化在时空尺度上的连接形成了脑功能网络。从网络的角度解析大脑神经信息处理机制是目前神经科学和控制科学等交叉学科研究的前沿热点之一。目前脑功能网络的研究主要是通过脑电图、脑磁图和磁共振成像等大尺度的信息采集技术解析不同脑区之间的功能连接关系,基于微电极阵列信息采集技术,以神经元为基本单元,空间精度更高,对于脑机制研究具有重要价值,但研究相对较少,许多问题尚待阐明。针对神经元锋电位(spike)发放序列,本文研究了神经元功能网络的构建方法,分析了其拓扑特性,并利用神经元功能网络解码了鸽子运动转向行为,验证了神经元功能网络在大脑神经信息处理机制解析中的有效性。主要研究内容和取得的成果如下:1)利用归一化互信息算法度量了spike序列之间的功能连接关系,构建了神经元功能网络,并结合Izhikevich神经元发放模型,验证了网络构建算法的性能。结果表明,基于互信息算法构建的神经元功能网络与真实网络之间的连接关系具有较高一致性,其性能优于传统的相关系数算法。2)分析了随机网络、规则网络和小世界网络等三种典型网络在不同节点连接概率、节点度和节点数量时的拓扑连接特性,并在此基础上分别研究了神经元spike发放率及神经元个数对网络拓扑特性的影响。结果表明,在一定范围内,随着spike发放率的增加,网络拓扑特性整体趋势呈抛物线形状;随着神经元数量的增加,网络拓扑特性整体呈现下降趋势。3)利用神经元功能网络特征,采用支持向量机神经解码算法,解码了十字迷宫目标导向任务中鸽子运动转向行为,并与基于spike发放率特征的解码算法进行了对比。结果表明,神经元功能网络具有较好的解码性能,无论是全局效率还是聚类系数特征,其解码正确率都优于神经元集群spike发放率特征的解码正确率。
[Abstract]:The brain network is one of the most complex networks in nature. Hundreds of millions of neurons in the brain form a large and complex network of brain structures. The synchronization of dynamic activities between clusters or brain regions forms a brain functional network on a space-time scale. Analyzing the neural information processing mechanism from the perspective of the network is the current crossover of neuroscience and control science. At present, the research of brain functional network is mainly through electroencephalogram (EEG). Large-scale information acquisition techniques such as magnetoencephalogram and magnetic resonance imaging are used to analyze the functional connections between different brain regions. Based on the technology of microelectrode array information acquisition, the neuron is taken as the basic unit and the spatial accuracy is higher. It is of great value to study the brain mechanism, but the research is relatively few, and many problems need to be clarified. In this paper, we study the construction method of neural functional network and analyze its topological characteristics, aiming at the neuronal spike-spike-releasing sequence. And the pigeon movement turn behavior is decoded by neural function network. The validity of neural functional network in the analysis of neural information processing mechanism is verified. The main research contents and results are as follows: 1) the functional connection between spike sequences is measured by normalized mutual information algorithm. The neural functional network is constructed, and the performance of the network building algorithm is verified by combining the Izhikevich neuronal distribution model. The results show that, The connection relationship between neural functional network and real network based on mutual information algorithm has high consistency, and its performance is better than that of traditional correlation coefficient algorithm. 2) the stochastic network is analyzed. The topological connection characteristics of three typical networks, such as regular network and small-world network, under different node connection probability, node degree and number of nodes. On this basis, the effects of the distribution rate of neuron spike and the number of neurons on the network topology are studied respectively. The results show that, with the increase of spike distribution rate, the overall trend of network topology is parabola. With the increase of the number of neurons, the overall topological characteristics of the network show a downward trend. 3) by using the neural decoding algorithm of support vector machine (SVM), the pigeon turning behavior in the target orientation task of the cross maze is decoded. Compared with the decoding algorithm based on the spike transmit rate feature, the results show that the neural network has good decoding performance, both global efficiency and clustering coefficient feature. The decoding accuracy is better than the decoding accuracy of the characteristics of the spike distribution rate in the neuronal cluster.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【相似文献】