基于脑电多特征融合的癫痫发作预测方法  

A seizure prediction method based on EEG multi-feature fusion

作　　者：高云园[1,2] 高博 罗志增 孟明[1] 张建海[2] GAO Yun-yuan;GAO Bo;LUO Zhi-zeng;MENG Ming;ZHANG Jian-hai(College of Automation,Hangzhou Dianzi University,Hangzhou 310018,China;Zhejiang Key Laboratory of Brain Computer Collaborative Intelligence,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学自动化学院,杭州310018 [2]浙江省脑机协同智能重点实验室,杭州310018

出　　处：《控制与决策》2023年第1期161-170,共10页Control and Decision

基　　金：国家自然科学基金项目(61971168,61871427,62171171);国家自然科学基金重点项目(U20B2074);之江实验室开放项目(2021MC0AB04)。

摘　　要：癫痫的发作会给患者的身体和精神造成极大的创伤,对癫痫发作的准确预测可以及时协助医生对患者采取治疗措施.为了准确预测癫痫发作,提出脑电特征和多通道脑电交互特征相融合的癫痫发作预测方法.首先,提出多尺度符号化排列传递熵对多通道脑电信号交互信息进行分析,生成同步矩阵,并通过显著性分析筛选与癫痫发作相关的重要脑电通道,减少不必要特征对分类的干扰;然后,对筛选通道后的脑电信号生成表征脑电信号特征的功率谱密度能量图(PSDED)和描述脑通道交互特征的同步矩阵图(SMD),将两个特征图融合,采用深度卷积神经网络(DCNN)对癫痫患者脑电信号进行分类识别,提高学习能力和泛化能力,分类准确率可达到96.825%;最后,在分类的基础上采用预测评价系统对癫痫发作预测性能进行评估,癫痫发作预测范围(SPH)为10 min和发作发生期(SOP)为10 min时,预测敏感性达到96.66%,误检率可达到0.03/h;当SPH为30min,SOP为10 min时,预测敏感性达到93.17%,误检率可达到0.05/h.与现有研究结果相比较,所提出方法具有较好的预测敏感度和较低的误检率.Epilepsy seizures can cause great physical and mental trauma to patients,and accurate prediction of epileptic seizures can assist doctors in treating patients in time.In order to accurately predict seizures,this paper proposes a seizure prediction method that integrates electroencephalogram(EEG) features and multichannel EEG interaction features.Firstly,the multi-scale symbolic alignment transfer entropy is proposed to analyze the multi-channel EEG signal interaction information,generate the synchronization matrix,and screen the important EEG channels related to seizures by significance analysis to reduce the interference of unnecessary features to the classification.Then,the power spectral density energy diagram(PSDED),which characterizes the EEG signal,and the synchronization matrix diagram(SMD),which describes the interaction characteristics of brain channels,are generated for the EEG signals after screening the channels.The two feature maps are fused,and the deep convolutional neural network(DCNN) is used to classify and identify the EEG signals of epilepsy patients,which improves the learning ability and generalization ability,and the classification accuracy can reach 96.825%.Finally,a prediction evaluation system is used on the basis of classification to evaluate the seizure prediction performance,and the prediction sensitivity reaches 96.66% and the false detection rate can reach 0.03/h when the seizure prediction range(SPH) is 10 minutes and the seizure onset period(SOP) is 10 minutes.When the SPH is 30 minutes and the SOP is 10 minutes,the prediction sensitivity reaches 93.17% and the false detection rate can reach 0.05/h.Compared with the results of existing studies,the proposed method has better prediction sensitivity and lower false detection rate.

关 键 词：脑电信号 癫痫发作预测 同步性分析 深度卷积神经网络 特征图融合 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]