一种基于稀疏表示模型的脑电图信号分析方法

Electroencephalogram signal analysis based on a sparse representation model

目的:癫痫是以脑内神经元异常放电致部分或整体脑功能障碍为特征的慢性疾患,模拟生物视觉感知系统,根据神经元响应的稀疏特性,对癫痫高危人群进行神经系统电生理筛查,以便及早发现和对相关人群进行干预。方法:选取适合的稀疏分解的匹配追踪算法,用新的较少的原子来重建正常的脑电信号和特定疾病类型的脑电信号,便于对各种神经系统疾病的脑电信号的特征波进行识别和提取。结果:处理16导标准脑电信号,分离出癫痫特征波,并对特征波进行识别,从而得到对癫痫的诊断,在此基础上将癫痫特征波反映射到16导标准电极,应用相关源电位软件对癫痫灶进行初步定位。结论:应用稀疏表示模型可以获取对脑电图信号的有效表示方法,通过对脑电图信号各分量进行有效的机器识别,归纳出系列特征波图谱,供临床诊断参考,从而降低了癫痫信号识别的工作量,提高了识别效率和正确率,实现癫痫的规模筛查。

AIM： Epilepsy is a chronic disease characterized by partial or overall brain disorder caused by the neuron paradoxical discharge in brain. This study simulated biological visual perception system for nervous system electrophysiological screening according to the sparse neuronal response characteristics of the high-risk population with epilepsy, so as to early detect and intervene the relevant population. METHODS： Using suitable sparse matching pursuit algorithm, normal electroencephalogram （EEG） and EEG of specific types of diseases were rebuilt with new less atom to identify and extract the characteristics of EEG in various nervous system diseases. RESULTS： After the treatment of 16-standard EEG, characteristic wave of epilepsy was isolated and identified, and the diagnosis of epilepsy was obtained. The characteristic wave reflection of epilepsy was stroke into the 16-standard electrode to preliminarily locate the epileptic foci using relevant sources of electric potential software. CONCLUSION： Sparse representation model can obtain EEG signals. Through the effective component machine identification of EEG, the series characteristic wave maps are summed up, which provide clinical diagnostic information, reduce the workload of epilepsy signal recognition, enhance the efficiency and accuracy of identification and realize scale screening of epilepsy.