经验模态分解和Kolmogorov测度的癫痫预测算法

针对头皮脑电信噪比低的缺点,提出了一种新的癫痫发作预测算法.首先对头皮脑电进行经验模态分解,去除伪差,保留包含主要癫痫预测信息的固有模态分量,然后用Kolmogorov测度来反映大脑的非线性动力学特征变化,并发现在癫痫发作之前,仅位于病灶区域附近导联的Kolmog-orov测度明显降低.通过对3例癫痫病人共5段长程头皮脑电信号的分析表明,这3例病人的平均发作预测时间为338 s,敏感性为66.7%,特异性为19.2%,因此该算法具有良好的临床应用前景.

Epileptic Seizure Prediction Method Based on Empirical Mode Decomposition and Kolmogorov Complexity

Aiming at the deficiency of scalp electroencephalogram(EEG),a new epileptic seizure prediction method is proposed,where the empirical mode decomposition(EMD) is utilized to remove the artifacts in scalp EEG signals,and the intrinsic mode functions containing the essential information to epileptic seizure prediction is retained.Then,Kolmogorov complexity is employed to reflect the non-linear dynamic characteristics of brains,which reveals that only the Kolmogorov complexity of electrodes near the epileptogenic area decreases significantly before seizures.The algorithm is validated by the clinical data collected from 3 epilepsy patients.The results show that the average prediction period gets 338 seconds,the mean sensitivity reaches to 66.7% and the specificity 19.2%.