基于脑电棘波频次的癫痫发作预测算法

Seizure Prediction Algorithm Based on Spike Rate in EEG

癫痫发作预测是近年来在神经科学领域中备受关注的课题。预测癫痫发作可以使医护人员或患者提前采取有效措施来预防和控制癫痫发作,在临床上具有重要意义。棘波是最基本的阵发性异常脑电活动,在分析和统计癫痫发作前期和发作期棘波频次不同表现的基础上,首次提出一种基于脑电棘波频次的癫痫预测算法。对脑电进行滤波以去掉高频干扰后,采用形态学滤波器检测脑电棘波数目,并计算各段脑电中棘波出现的频次,最后根据棘波频次的变化预测癫痫的发作。采用本算法对21例癫痫患者长程颅内脑电进行癫痫预测,准确率达到74.7%,每小时错误预测次数仅为0.111次。结果表明,所提出算法能够有效地预测癫痫发作。

Seizure prediction is a field of great interest in neuroscience communities, because it can make doctors or patients to take effective measures to prevent and control seizures. Spike is the basic paroxysmal EEG activity. In this work the different performance of Spike Rate （SR） was analyzed in preietal and ictal EEG. An algorithm based on SR detection by morphological filtering to predict seizure was proposed and investigated. First, the high frequency artifacts of EEG were removed by filtering. Then, the improved morphological filter was used to detect the spike number and calculate SR （Spike Rate） of every segment of EEG. Finally, seizures were predicted according to the variation of SR. The proposed algorithm was evaluated with long-term EEGs of 21 patients with epilepsy in the experiment, and the sensitivity reached 74.7% with a false prediction rate of 0. 1 l l/h. The results show that the proposed algorithm can predict seizures effectively.