目的 癫痫是由多种病因引起的慢性脑功能障碍综合征,及时的发作预报,对于建立新的治疗方法和改善患者的生活质量有着至关重要的作用.目前大部分脑电分析算法存在计算速度慢、适应性差等问题,无法满足癫痫脑电发作预报的要求.方法 本文应用自回归模型对脑电信号进行特征提取,支持向量机(support vector machine,SVM)作脑电各个时期分类器,并与Lempel-Ziv复杂度分析计算相结合,准确识别发作前期,以实现癫痫的发作预报.结果 应用弗莱堡大学数据对上述方法的有效性进行验证.仿真结果表明,该方法得到的发作漏检率、误报率较低,预报提前时间较长.结论 将AR模型和Lempel-Ziv复杂度相结合,对癫痫发作预报的实现,有一定参考价值和意义.%Objective Epilepsy is a chronic brain dysfunction syndrome caused by many diseases. The predictions of epilepsy seizure are significant for both the establishment of new treatment methods and the improvement of the patients' life qualities. The current EEG analysis algorithm cannot meet the requirement of epileptic seizure prediction for the slow computation and the poor adaptability. Methods This paper applies autoregressive( AR ) model for feature extraction, a support vector machine as a classifier, and combines Lempel-Ziv complexity( LZC ) to identify preictal accurately. Results Using the data from Freiburg University, the simulation results show that the methods used in this paper achieve a lower false alarm rate, a lower failed reporting rate and a longer lead time. Conclusions This paper provides references for the realization of the epileptic seizure prediction by combining AR model and LZC.
展开▼
