摘要
本研究提出基于EEG序列模糊相似性指数方法预测癫痫发作。首先,结合复自相关法和Cao法对EEG序列进行了相空间重构;然后,计算相关积分时用Gaussian函数代替Heavyside函数,克服了Heavyside函数的刚性边界问题,使得计算相似性指数更加准确和可靠;最后,分析大鼠癫痫EEG信号,检测癫痫发作前期状态。分析结果表明模糊相似性指数方法能够比动态相似性指数方法获得更长的预测时间和更低的错误预测率。
This paper proposes a fuzzy similarity method to predict epileptic seizures with electroencephalography (EEG). First, muhiple-autocorrelation and Cao's method are employed to reconstruct a phase space of EEG recordings. Second, instead of Heavyside function is Gaussian function used in correlation integral for calculating a similarity index, so the crisp boundary of the Heavyside function is eliminated to make the similarity index is more accurate and reliable. Finally, the fuzzy similarity index is applied to indicate the preictal state of nine rats with EEG signals. The result shows that the fuzzy similarity index is better than dynamical similarity index in increasing anticipation time and decreasing false prediction rate for the prediction of epileptic seizure.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2006年第3期346-350,381,共6页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金资助项目(60575012)
关键词
EEG信号
模糊相似性指数
癫痫发作
预测
相空间重构
EEG signals
fuzzy similarity index
epileptic seizure
prediction
phase space reconstruction