基于复杂网络的癫痫脑电分类与分析
Classification and analysis of epileptic EEG based on complex networks

摘要： 为提取癫痫发作与间歇期脑电信号的特征,提出利用构建癫痫EEG(electroencephalogram)网络的方法来刻画脑电信号。研究各变量均可测情况下的Lorenz和R？ssler混沌系统,利用其各变量的输出混沌时间序列构建复杂网络,发现构建的复杂网络拓扑图与其混沌吸引子存在形态相似性,说明由时间序列构建的复杂网络能刻画其原信号特征。对于多维系统中仅有一维可测时,多维时间序列由相空间重构得到。利用相空间重构方法对癫痫发作和间歇期脑电信号构建复杂网络进行分析。研究结果表明,癫痫发作时其网络拓扑较间歇期存在明显不同,且其平均路径长度显著增加,而递归率及其波动范围都显著降低,这些网络特性可以用来刻画脑电信号的特征,从而为癫痫疾病的自动辨识与预测提供基础。
Abstract: To extract epileptic EEG features in the ictal and interictal period, a method of depicting epileptic EEG was proposed by transforming epileptic EEG time series to epileptic networks. Chaotic multi-dimensional time series coming from the Lorenz system and R？ssler system were used to construct a complex network,in which all the variables could be measured. It was found that there was morphological similarity between topology of the complex networks and the attractor of chaotic system. This indicated that complex networks constructed from time series could depict the characteristics of the original signals. For only one measureable variable, multi-dimensional time series were obtained by reconstruction of the phase space. Therefore, the epileptic EEG network was constructed and analyzed in the ictal and interictal period. The results showed that epileptic EEG network topologies in the ictal period were significantly different from that in the interictal period. Meanwhile, the average path length of the network increased significantly and recurrence rates decreased significantly in the ictal period comparing to in the interictal period. These network features could be used to depict the characteristics of EEG time series and could provide the basis for epilepsy automatic identification and prediction