基于支持向量机的多通道癫痫发作预测

癫痫是一种大脑神经系统疾病,具有突发性和反复性,对患者的生命安全构成极大的威胁,有效预测癫痫对该病的预防和治疗具有重要的意义。为此,提取来自德国弗莱堡大学癫痫预测中心21个病人的公开数据集。利用独立成分分析方法对原始数据进行去冗余操作,自回归模型被用来对癫痫脑电进行特征提取。支持向量机模型和滤波器将预测问题转化为二分类问题。蒙特卡洛统计方法使得最终的结果具有统计学上的意义。实验结果表明,该模型能够提前30 min~70 min预测到癫痫的发生,且误报率将近0,能为临床癫痫预警系统提供较好的理论依据。

Multi-channel Seizure Prediction Based on Support Vector Machine

Epilepsy is a brain disease. As the disease is sudden and repeated, which poses a great threat to safety of patients, effective prediction to seizure is of important significance to prevention and treatment. In this paper, dataset comes from University of Freiburg, Germany Prediction Center. Independent Component Analysis（ICA） is used to remove redundancy. Auto regression model is used to extract multi-channel features of changing trend along with time series. Prediction is transferred to classification by Support Vector Machine（SVM） and filter. All the results can be finally got by Monte Carlo statistical methods. Results show that the models can predict seizures in advance 30 min-70 rain with false positive rate nearly zero, which may provide good theoretical basic for developing clinical epilepsy warning system.