自动检测儿童脑电中癫痫波的方法研究  

Study of Automatic Detection of Epileptiform Waves in Children EEG

作　　者：李莹[1] 欧阳楷[1] 

机构地区：[1]首都医科大学生物医学工程学院,北京100054

出　　处：《中国生物医学工程学报》2005年第5期541-545,共5页Chinese Journal of Biomedical Engineering

摘　　要：研究的是儿童脑电数据中癫痫波的检测算法。儿童脑电比成人脑电复杂,因而检测算法的设计与成人脑电的检测算法有所不同。综合考虑了国内外多种研究方法,确定了一种以半波检测方法为基础的检测算法。大体思想是:通过初步筛选,找出符合棘、尖波特征的脑电波,然后使用专家规则剔除符合这些特征的几种常见伪迹。再对选出的癫痫特征波利用模式识别的聚类思想分为三类,确定出更为可能是阳性波的特征,对脑电数据作进一步筛选。然后,用小波分析对检出的波再处理,去除不易被识别的假阳性波。最后使用人工神经网络对检出的癫痫波分类。Most of the EEG data collected in this paper for detecting epileptiform waves are from children, and infants. Children＇s EEG is more complex than that of adults＇. By reviewing the reference papers in this field, A detection method based on a means of ＇half-wave＇ detection was proposed. The detection method included five steps. First, the EEG data was roughly filtered by a threshold, the waves whose characteristics were coincident with the conditions are seleoted. Second, using expert experience including most of the false-positive detection to get rid of artifacts. In the third step, the remanent waves were divided into three classes by cluster method, then the remarkable epileptiform waves were clustered into one class. The relationship among the three classes was analyzed in this work, the waves not distinct enough to be recognized by doctors were removed off. In the fourth step, further detection with wavelet analysis was conducted. At last, artificial neural networks （Bp）were used to classify all the recognized waves into four class：spike,sharp,spike ＆ slow,sharp ＆ slow.

关 键 词：脑电 癫痫波 自动检测 人工神经网络 

分 类 号：R318.04[医药卫生—生物医学工程]