基于自适应多尺度脑功能连接的局灶性癫痫发作检测方法研究  

Seizure Detection in Focal Epileptic Patients Based on Adaptive Multi-Scale Brain Functional Connectivity

作　　者：徐嘉阳 杨婷婷 李雯 李扩[2] 杜昌旺[2] 刘晓芳[2] 盛多铮 闫相国[1] 王刚[1] Xu Jiayang;Yang Tingting;Li Wen;Li Kuo;Du Changwang;Liu Xiaofang;Sheng Duozheng;Yan Xiangguo;Wang Gang(Institute of Health and Rehabilitation Science,School of Life Science and Technology,The Key Laboratory of Biomedical Information Engineering of Ministry of Education,Xi’an Jiaotong University,Xi'an 710049,China;Department of Neurosurgery,First Affiliated Hospital of Xi’an Jiaotong University,Xi'an 710061,China;Beijing Braincare Technology Co.Ltd,Beijing 100071,China)

机构地区：[1]生物医学信息工程教育部重点实验室,西安交通大学生命科学与技术学院,健康与康复科学研究所,西安710049 [2]西安交通大学第一附属医院神经外科,西安710061 [3]北京瑞尔唯康科技有限公司,北京100071

出　　处：《中国生物医学工程学报》2022年第4期393-401,共9页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金(32071372;31571000);陕西省自然科学基础研究计划资助项目(2020JM-037)。

摘　　要：利用长时程脑电图检测癫痫发作是临床中较为广泛的应用,然而这项工作乏味、耗时,且很大程度上依赖于临床医生的自身经验和主观判断,准确性和可重复性也较低。针对长时程脑电图检测癫痫中存在的问题,提出一种基于自适应多尺度脑功能连接的癫痫发作检测方法(AMBFC),并选取10例局灶性癫痫患者的发作期和非发作期的样本作为研究对象。首先在一个滑动时间窗内,通过多元经验模态分解(MEMD)提取19通道脑电信号的7个本征模函数(IMF)分量及残差;然后建立多变量自回归(MVAR)模型,利用有向传递函数(DTF)提取流出信息强度,进行特征组合,并通过主成分分析(PCA)降维保留原始特征数目的85%;最后经代价敏感支持向量机(CSVM)分类区分发作期和非发作期脑电,并通过五重交叉验证进行癫痫发作检测算法的效果评价。结果表明,AMBFC算法检测脑电癫痫发作得到的平均准确率为98.6%,精确率为81.9%,召回率为81.4%,F2值为0.80。与各IMF分量、DTF-CSVM算法等检测结果相比,AMBFC算法更具有优越性。有望应用于长时程脑电的实时监测。Long-term EEG has been widely used to detect epileptic seizures in clinical practice. However, this approach is tedious and time-consuming, and largely depends on clinicians′ experience and subjective judgment. As a result, the accuracy and repeatability of the manual detection results are low. In this study, with the aim of solving the problem by using long-term EEG to monitor epileptic seizures, we proposed a novel adaptive and multiscale brain functional connectivity(AMBFC) method for epilepsy detection. Samples of 10 epilepsy patients during the seizure and non-seizure periods were selected as the research subjects. First, within a sliding time window, seven IMF components and residuals of the 19-channel EEG signal were extracted by MEMD. Then MVAR model was established to extract the outflow information intensity by the directional transfer function, and the features were combined and dimensionally reduced using PCA. Finally, the CSVM model was used to classify the seizure phase and non-seizure phase EEG, and the effect of epileptic seizures was evaluated through five-fold cross-validation. The results showed that the average accuracy rate of AMBFC algorithm for detecting EEG seizures was 98.6%, accuracy rate was 81.9%, recall rate was 81.4%, and F-measure value was 0.80. Compared with the detection results of the different IMF components, DTF-CSVM algorithm and methods in recent literatures, AMBFC algorithm was better. Except for the high accuracy, the proposed algorithm also achieved high precision, recall and F2 value. In conclusion, this method can be applied to real-time monitoring of long-term EEG.

关 键 词：脑电信号 癫痫发作 多元经验模态分解(MEMD) 有向传递函数(DTF) 代价敏感支持向量机(CSVM) 

分 类 号：R318[医药卫生—生物医学工程]