脑电信号的分形截距特征分析及在癫痫检测中的应用  

Fractal Intercept Analysis of EEG and its Application for Seizure Detection

作　　者：王玉[1] 周卫东[1] 李淑芳[1] 袁琦[1] 耿淑娟[1] 

机构地区：[1]山东大学信息科学与工程学院,济南250100

出　　处：《中国生物医学工程学报》2011年第4期562-566,共5页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金(30870666);山东省科技发展计划项目(2010GSF10243);山东大学自主创新基金(2009JC004)

摘　　要：脑电信号的非线性特征会随癫痫发作而改变,脑电信号的特征分析和检测对癫痫的诊断和治疗具有重要意义。提出对癫痫脑电信号进行毯子维和分形截距的特征分析,并将分形截距应用于癫痫脑电信号的检测。首先提取脑电信号的分形截距和毯子维特征,并对两种特征的均值和方差进行比较,最后使用支持向量机分类器,实现脑电信号的分类检测。发现癫痫发作时脑电信号的分形截距显著高于发作间期,而脑电信号的毯子维在发作前后变化规律则不明显。将分形截距作为分类特征,能有效地区分癫痫脑电与间歇期脑电,具有较强的癫痫脑电检测性能,分类检测的准确率达到96%以上。Nonlinear features of electroencephalogram（EEG） vary with epileptic seizure,and the feature analysis and detection of epileptic EEG are significant in diagnosis and therapy of epilepsy.This paper presents an epileptic EEG analysis approach based on blanket dimension and fractal intercept features,and applies fractal intercept to epileptic EEG detection.We extract fractal intercept and blanket dimension features of EEG,and compare the mean and variance of those two features.Then,a support vector machine is applied to classify epileptic EEG signals.It is found that the fractal intercept features of EEG during epileptic seizure are significantly higher than interictal EEG＇s,while the blanket dimension features of EEG show no significant differences before and after seizures.The fractal intercept as a classification feature could be used to distinguish epileptic EEG from interical EEG with high performance for seizure detection,and the classification accuracy is up to 96%.

关 键 词：脑电 癫痫 非线性特征 毯子维 分形截距 

分 类 号：R318[医药卫生—生物医学工程]