基于同步似然的gamma子带功能网络构建与鸽子转向行为解码

发布时间：2018-02-25 10:23

本文关键词： 目标导向抉择行为 gamma子带同步似然分析拓扑特性 LOO-kNN解码算法　出处：《郑州大学》2017年硕士论文　论文类型：学位论文

【摘要】：脑是自然界中最复杂的网络,脑中数以亿计的神经元之间不同的连接模式不仅编码了动物的行为,而且控制着躯体的功能输出。因此,与特定行为和功能相对应的神经元网络信息处理机制的解析,目前已成为神经科学和控制科学等交叉领域新的研究热点,有助于深化对大脑特定功能和机制的理解。本文针对动物目标导向抉择行为的神经信息解码问题,以鸽子为模式动物,首先通过植入式微电极阵列记录了鸽子弓状皮质尾外侧区神经信号,根据转向过程中局部场电位(local field potential,LFP)不同频带的时频特性变化,确定了转向行为的编码频段;然后利用同步似然算法构建了gamma子带功能网络,并分析了网络的拓扑特性,确定了转向行为的解码时间窗;最后提取了gamma子带功能网络特征,并利用联合留一法(leave one out,LOO)和k近邻(k-nearest neightor,kNN)的神经信息解码算法解码了鸽子的转向行为。本文已完成的工作和取得的研究成果概括如下:1)利用多窗口重叠功率谱估计和小波变换技术,从时域、频域、时频域三个角度对鸽子转向过程中的LFP信号进行了特性分析,以确定与鸽子转向行为相关的LFP信号特征频带。结果发现,与等待区相比,转向区的gamma子带(40~60 Hz)能量显著增加,而其它子带并没有明显变化,这表明LFP信号gamma子带编码了鸽子的转向行为。2)利用同步似然分析算法,构建了LFP信号gamma子带功能网络,度量并分析了网络的拓扑特性。对比分析了等待区与转向区、不同方向以及转向前后gamma子带功能网络的拓扑特性,发现转向区的网络聚类系数和全局效率显著高于等待区,而且转向区不同方向的网络特性之间的差异也比较明显。进一步的研究发现转向后聚类系数和全局效率显著高于转向前,这表明转向后gamma子带功能网络包含了鸽子的转向行为信息。3)利用网络的连接强度值和主成分降维技术,提取了gamma子带功能网络特征,并利用LOO-kNN解码算法解码了鸽子的转向行为,对比分析了网络特征和能量特征的解码正确率。结果表明,网络特征的解码正确率(0.74±0.08)显著高于能量特征的解码正确率(0.61±0.12)。同时,通过对转向过程中gamma子带网络特征解码正确率的动态特性分析发现,不同鸽子的解码正确率峰值大都出现在转向后,这表明在鸽子转向过程中目标可能起到了关键作用。
[Abstract]:The brain is the most complex network in nature. The different patterns of connections between hundreds of millions of neurons in the brain not only encode the behavior of animals, but also control the functional output of the body. The analysis of neural network information processing mechanism corresponding to specific behavior and function has become a new research hotspot in the intersecting fields of neuroscience and control science. It is helpful to deepen the understanding of the specific functions and mechanisms of the brain. In this paper, the pigeon is used as a model animal to solve the problem of neural information decoding for the goal-oriented behavior of animals. Firstly, the neural signals in the lateral caudal region of the arcuate cortex of pigeons were recorded by implanted microelectrode arrays. According to the time-frequency characteristics of different frequency bands of local field potential (LFP), the coded frequency band of the steering behavior was determined. Then the gamma sub-band functional network is constructed by using synchronous likelihood algorithm, and the topological characteristics of the network are analyzed, and the decoding time window of the steering behavior is determined. Finally, the feature of the gamma sub-band functional network is extracted. And the neural information decoding algorithm based on joint leave one outloo) and k-nearest neighbor KNN) is used to decode the dove's turn behavior. The work done and the research results obtained in this paper are summarized as follows: 1) using multi-window overlapping power spectrum estimation. And wavelet transform technology, The characteristics of LFP signal in pigeon steering are analyzed from three aspects: time domain, frequency domain and time frequency domain, in order to determine the characteristic frequency band of LFP signal related to dove steering behavior. The energy of the gamma subband 4060 Hz in the steering region increased significantly, but the other subbands did not change significantly. This indicates that the LFP signal gamma subband encodes the dove turning behavior. 2) using the synchronous likelihood analysis algorithm, the LFP signal gamma subband functional network is constructed. The topological characteristics of the network are measured and analyzed, and the topological characteristics of the waiting area and the steering area, the different directions and the gamma subband function network before and after steering are compared. It is found that the network clustering coefficient and global efficiency of the steering region are significantly higher than that of the waiting area. Moreover, the difference between the network characteristics of different directions in the steering region is obvious. Further studies show that the post-steering clustering coefficient and global efficiency are significantly higher than those before turning. This indicates that the post-steering gamma subband functional network contains dove turning behavior information. 3) the feature of gamma subband functional network is extracted by using the connection strength value and principal component dimensionality reduction technique of the network, and the pigeon's turn behavior is decoded by LOO-kNN decoding algorithm. The results show that the decoding accuracy of network features is 0.74 卤0.08, which is significantly higher than that of energy features (0.61 卤0.12). By analyzing the dynamic characteristics of the decoding accuracy rate of gamma subband network in the process of steering, it is found that the peak decoding accuracy rate of different pigeons occurs after the turn, which indicates that the target may play a key role in the process of pigeon steering.
【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】