基于SVM算法癫痫脑电的研究初探

Preliminary research of epilepsy brain electricity based on SVM algorithm

目的:通过算法的选择,对癫痫患者进行无创性脑电图筛查,以期及早发现、正确干预具有癫痫发病倾向的人群,有效降低癫痫的发病率、致残率和死亡率。方法:通过对基于支持向量机分析技术的数字脑电图(EEG)信号正常和癫痫特征波的提取、自动识别和分型等研究,以期实现癫痫的自动、规模性筛查。结果:癫痫患者和健康人的16导脑电信号在能量特征上表现出较高的模式可分性,支持向量机作为一种新的机器学习方法,具有较强的泛化能力,基于该算法的EEG信号的分类方法会成为诊断隐匿癫痫患者的一种新的可行途径。结论:SVM算法适合有限样本(小样本)问题,对自发脑电信号的分类可以取得较好的效果,可用于癫痫脑电信号异常的筛选。

Objective To select algorithm for noninvasive EEG screening of epilepsy patients with a view to early detection and reduction of the incidence of epilepsy, morbidity and mortality. Methods Electroencephalogram（EEG）signal characteristics of the normal and epilepsy wave were extracted, automatically identified and classified based on support vector machine （SVM） analysis with a view to achieving epilepsy automatic scale screening. Results The EEG characteristics energy displayed by the model between epilepsy patients and healthy people could be divided obviously. As a new machine learning methods, SVM had a strong ability to generalize. EEG signals based on the algorithm of the classification would become diagnosis of epilepsy patients misprision of a new viable avenue.Conclusion SVM is suitable for the limited samples （small samples）. The spontaneous EEG classification with SVM can achieve better results, so it can be used to epileptic EEG abnormality screening.