MF-DFA在癫痫发作期及发作强度检测中的应用  

Detection of Seizure Timecourse and Index of Seizure Intensity with Multifractal Detrended Fluctuation Analysis

作　　者：张翔[1,2] 丁勇[1] 刘小峰[1,2] 

机构地区：[1]河海大学物联网工程学院,常州213022 [2]河海大学一法国Alderbaran Robotics认知与机器人实验室,常州213022

出　　处：《数据采集与处理》2013年第5期664-671,共8页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(60905060,31101643)资助项目;中央高校基础研究(2011B11114,2012B07314)资助项目;国家大学生创新训练项目(201302941059G)资助项目;国家重大基础研究计划(“九七三”计划)(2010CB327902)资助项目

摘　　要：癫痫患者的脑电特征是临床诊断、分类和预报癫痫的重要依据。作为测度时间序列内在模式变化的近似熵和样本熵成为一种临床癫痫分类和发作预报的重要方法,由于受到序列长度、嵌入维数以及阈值设置的影响,难以准确检测序列内模式突变的时刻。为准确检测脑电癫痫样放电时刻及其强度,提出了一种癫痫发作及强度检测的多分形去趋势波动分析方法(MF-DFA),并与基于样本熵的癫痫放电检测作进一步比较分析。采用头皮表面脑电与颅内脑电临床数据做测试实验,结果表明:MF-DFA能够精确捕捉到发作时刻,并能够定量刻画癫痫发作强度。EEG features of epilepsy patients are an important index for the diagnosis and classification of epilepsy,as well as for long-term monitoring and predicting seizures.Both approximate entropy and sample entropy are used to extract features of epilepsy.However,both methods are sensitive to changes in their parameters,e.g.data points,embedding dimension,and tolerance.A new method based on multifractal detrended fluctuation analysis （MF-DFA）is presented,for detecting epileptic seizures and measuring seizure intensity.The performance of the new method is evaluated by either scalp EEG or intracranial EEG （EcoG）.Results show that it is capable of accurately capturing the early ictal onset of epileptic seizure and extracting the quantitative description of seizure intensity.

关 键 词：癫痫检测 样本熵 发作起始时刻 发作强度 多重分形去趋势波动分析 

分 类 号：R318[医药卫生—生物医学工程]