基于空频域特征分析方法的癫痫发作预测  

Epileptic seizure prediction based on spatial-frequency domain feature analysis

作　　者：韩凌[1] 王宏[2] 

机构地区：[1]东北大学中荷生物医学与信息工程学院,沈阳110819 [2]东北大学机械工程与自动化学院,沈阳110819

出　　处：《仪器仪表学报》2014年第11期2501-2507,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61071057)项目资助

摘　　要：癫痫发作具有突发性和反复性,对患者生命安全构成巨大威胁。为了对癫痫发作进行有效地预测,提出了空频域特征分析的癫痫发作预测方法。将多变量相位同步参数、希尔伯特边际谱和希尔伯特加权频率组成一个三维的特征向量作为空频域特征值,输入到支持相量机中,实现癫痫的发作预测,最后采用癫痫发作预测特征方法对预测结果进行评估。实验结果表明：采用空频域特征分析方法对δ波和θ波的癫痫发作预测,癫痫预测范围在30~45分钟,患者有足够的时间采取措施应对;癫痫发作周期在5~10分钟,缩短患者等待时间,降低焦虑程度;与多种相关方法进行比较,该方法具有较低的错误预报率和较高的预测敏感度。Epileptic seizure with sudden and repeatability poses a great threat to the patient. To effectively predict the epileptic seizure, a method based on the space-frequency domain feature analysis is proposed. A three dimensional feature vector composed ofmultivariate phase synchronization, Hilbert marginal spectrum and Hilbert weighted frequency compose as space-frequency domain feature values is inputinto a support vector machine（SVM） to predict epileptic seizure. The seizure prediction characteristic is used to assess the prediction performance. Experimental results show that the prediction horizon time for δ rhythm and θ rhythm is 30-45 minutes, and patients can have enough time to take measures to deal with seizures. Seizure occurrence period is 5-10 minutes, so that the waiting time for patient can be shortened and the anxiety of patient reduced. Compared with a variety of relevant methods, the proposed method has lower false prediction rate and higher prediction sensitivity.

关 键 词：希尔伯特黄变换 多变量相位同步分析 希尔伯特边际谱 希尔伯特加权频率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TH782[自动化与计算机技术—计算机科学与技术]