癫痫脑电特征波的综合检测分类方法研究

STUDY ON SYNTHESIZED MULTI-METHODS TO DETECT AND CLASSIFY THE EPILEPTIC WAVES IN THE EEG

本文将小波变换、人工神经网络、专家规则判据等多种检测方法有机地结合起来 ,用于癫痫脑电特征波的检测与分类 ,以充分发挥不同方法的优势。这种综合检测分类方法是先将预处理的多导脑电时间序列经小波变换将脑电中癫痫特征波在不同尺度下分离出来 ,再对选出的癫痫嫌疑波进行特征参数提取 ,然后把特征参数送入已经训练好的人工神经网络进行分类识别 ,最后再由专家规则判断筛选并作出检测分类统计报告。研究表明 ,该方法具有很好的信号特征提取和屏蔽随机噪声能力 ,获得了较好的检出率 ;尤其适合于非平稳、非线性生物医学信号的检测分类 。

Several signal processing methods, such as wavelet transforms, artificial neural networks and expert rules etc, are synthesized to detect and classify the epileptic waves in the EEG in order to sufficiently develop the advantages of different methods. In this synthesized multi method, the epileptic waves are detected from pre processed EEG data at different scales by wavelet transforms, then the characteristic parameters of the chosen candidates of epileptic waves are extracted and sent into the well trained neural networks to identify and classify the true epileptic waves. At last, the epileptic waves are detected by the expert rules and the statistic results of detection and classification are reported. It is realized that, the synthesized multi method has good ability to extract signal features and to shield the signals from the random noise. This method is especially benefited to the analysis of the biomedical signals, which are usually non placid and non linear.