基于功率谱及有限穿越可视图的癫痫脑电信号分析算法

发布时间：2018-06-22 20:53

本文选题：脑电信号 + 癫痫　；参考：《计算机应用》2017年01期

【摘要】：针对可视图(VG)算法存在噪声鲁棒性差的问题,提出一种改进的有限穿越可视图(LPVG)建网方法。该算法基于可视图(VG)算法的可视性准则,并设定有限穿越视距,将时间序列中满足条件的点连接起来,从而将时间序列映射为网络。首先,对LPVG算法进行性能分析;然后,将LPVG算法结合功率谱密度(PSD)算法应用到癫痫发作前、中、后脑电信号的识别上;最后,提取三种状态下癫痫脑电信号的LPVG网络特征参数,研究癫痫对网络拓扑结构的影响。仿真结果表明,与VG和水平穿越可视图(HVG)相比,虽然LPVG算法的时间复杂度较高,但是LPVG对信号中的噪声具有较强的鲁棒性:分别对周期、随机、分形和混沌四种时间序列进行LPVG建网,发现随着噪声强度增大,LPVG网络聚类系数的波动率均为最低,分别为6.73%、0.05%、0.99%和3.20%。接下来对脑电信号的PSD和LPVG建网分析结果表明,癫痫发作中,PSD值在delta频带下显著增强,而在theta频带下显著降低;LPVG网络拓扑结构有所改变,网络中各模块的独立性有所提高,网络的平均路径长度增大,复杂度降低。所提的功率谱密度和有限穿越可视图算法能够有效表征癫痫前、中、后三种状态下的脑电信号能量分布和单通道信号可视化后的网络拓扑结构的异常,为癫痫的病理研究和临床诊断提供帮助。
[Abstract]:Aiming at the problem of poor noise robustness in visual graph (VG) algorithm, an improved finite traversal visual graph (LPVG) network method is proposed. The algorithm is based on the visibility criterion of the visual graph (VG) algorithm, and sets the limited traversing horizon to connect the points in the time series which satisfy the conditions, so that the time series can be mapped to the network. Firstly, the performance of LPVG algorithm is analyzed. Then, the LPVG algorithm combined with power spectral density (PSD) algorithm is applied to the recognition of EEG signals before, during and after epileptic seizures. Finally, the LPVG network characteristic parameters of epileptic EEG signals in three states are extracted. The effect of epilepsy on network topology was studied. The simulation results show that, compared with VG and HVG, LPVG has higher time complexity, but LPVG is robust to noise in the signal: periodic and random, respectively. Fractal and chaotic time series are used to construct LPVG network. It is found that the fluctuation rate of clustering coefficient of LPVG network is the lowest with the increase of noise intensity, which is 6.730.05% and 3.20%, respectively. Then, the results of delta and LPVG network analysis showed that the PSD value of epileptic seizures was significantly enhanced in delta frequency band, while the topological structure of LPVG network was significantly decreased under theta band, and the independence of each module in the network was improved. The average path length of the network increases and the complexity decreases. The proposed power spectral density and finite traversal visual graph algorithm can effectively characterize the energy distribution of EEG signals before, during and after epilepsy and the anomalies of network topology after visualization of single channel signals. To provide help for pathological study and clinical diagnosis of epilepsy.
【作者单位】：天津职业技术师范大学信息技术工程学院;唐山市工人医院神经内科;天津大学电气与自动化工程学院;
【基金】：国家自然科学基金资助项目(61302002,61601331) 天津市自然科学基金资助项目(14JCYBJC15400,14JCQNJC01200) 唐山市科技支撑项目(14130223B) 天津职业技术师范大学预研项目(KYQD1611)~~
【分类号】：R742.1;TN911.6

【相似文献】