摘要
脑电癫痫波的自动检测与分类是具有重要临床意义的课题。现存的算法大都着重于对棘、尖波形的检测 ,而忽略了慢波所包含的有用信息。为满足临床要求 ,论文提出了一种改进的脑电癫痫波自动分析系统。系统采用“分层次、多方法”的检测策略 ,兼顾了各种癫痫病理波形 ;整个处理过程综合应用了自适应预测、小波变换、人工神经网络、启发式规则等多种信号处理方法。经临床数据测试 ,该系统对癫痫波的总检测率达 83.6 % ,误检率为 1.1%。通过分层次处理 ,运用多方法的结合 ,可以提高检测敏感度和特异度 ,减少计算量 。
Most automatic epilepsy detection systems focus on sharp transients although slow waves also provide much diagnostic information. Therefore, a new system was developed that analyzes all kinds of waveforms, especially slow waves and slow wave based activities. The system processes the signal in several hierarchical stages and utilizes various digital signal processing (DSP) methods including adaptive prediction, wavelet transform, artificial neural network and heuristic rules. Test on clinical recordings indicates the system has a detection rate of 83.6% and a false detection rate of around 1.1%. The hierarchical multi method strategy improves the system sensitivity and selectivity while greatly reducing the computational load, so it can fulfil the need for long term electroencephalography (EEG) data processing.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2002年第3期304-308,共5页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目 (3 9670 2 12 )
关键词
脑电癫痫波
自动分析系统
自适应预测
小波变换
人工神经网络
epileptic electroencephalography (EEG)
automatic detection
adaptive prediction
wavelet transform
artificial neural network
heuristic rules
