用脑电双谱分析和人工神经网络预测癫痫发作的研究

目的：根据癫痫患者脑电信号具有非高斯非线性随机特性，应用高阶累积量技术对癫痫患者的脑电信号进行双谱估计，进而结合神经网络研究发作前脑电对癫痫发作预报的价值，以寻求更加敏感和准确的发作预报参量和临床监护方法。方法：对7例癫痫患者在不同发作阶段特别是发作前夕的八导脑电信号进行双谱估计，提取各导脑电的双相关指数和加权双谱权重中心，研究了在不同发病阶段的脑电信号的高斯偏离程度，使用一个四层(24-10-2-1)的神经网络实现分类。神经网络的训练和测试采用去一循环对比法。结果：不同发作阶段时癫痫脑电信号的高斯偏离程度明显不同，其中双相关系数能够敏感区分癫痫的不同发作阶段；双相关系数和加权双谱中心作为人工神经网络输入时可提前12-24s预报癫痫的发作。结论：双谱分析、双相关系数为癫痫脑电信号的研究提供了一些新的思路，有望成为临床监护预报癫痫发作的一个指标。

Prediction of epileptic seizures using bispectrum analysis of electroencephalograms and artificial neural network

Objective:Based on the fact that the signals of epileptic electroencephalogram(EEG) possess non-Gaussian and nonlinear stochastic properties,the higher-order statistic methods are used to estimate the bispectrum of epileptic EEG,in order to obtain more sensitive and accurate parameters for clinical epileptic monitoring and seizure prediction.Method: This paper presented a new approach, based on bispectrum analysis of EEGs and artificial neural network(ANN), which predicted seizures in seven patients with epilepsy. Eight channels of EEG were collected in each subject in Epilepsy Center of Xijing Hospital.The Guassian deviations of the EEGs in different stages of seizure were studied,The bicoherence index and the weighted center of EEG bispectrum(WCOB) were extracted from the EEG bispectrum contour and a four layer(24-10-2-1) ANN was employed for prediction. Training and testing the ANN used the'leave one out'method.Result:Obvious difference of Guassian deviations in different stages of seizure existed and the bicoherence coefficient could discriminate the different seizure stages sensitively.The proposed system was able to correctly predict the succedent seizures and prediction times were from 12 to 24 seconds, prior to the onset of epileptic seizures.Conclusion:Bispectrum estimation and analysis of bicoherence coefficient can offer some new ideas for research of epileptic EEG and maybe they are potential parameters for clinical epileptic monitoring.