基于aEEG尖峰和cEEG棘波提取的癫痫发作检测算法

Epileptic seizure detection algorithm based on spike extraction of aEEG and cEEG

针对癫痫发作检测方法中数据量庞大导致模型训练时间过长、检测结果与临床医生标记结果存在差异等问题,提出了一种基于振幅整合脑电图(amplitude integrated electroencephalography,aEEG)尖峰和连续脑电图(continuous electroencephalography,cEEG)棘波提取的癫痫发作检测算法。此发作检测算法以MATLAB软件为基础开发平台,利用形态学滤波技术对多通道aEEG脑电信号进行尖峰提取,确定癫痫发作所在通道。然后对该通道cEEG进行形态学滤波,并计算每个5 s长脑电片段的棘波率(spike rate,SR),使用阈值法进行癫痫发作检测,如果某个脑电片段的SR大于预设的发作阈值,则说明该片段为癫痫发作片段。实验结果表明:该算法的癫痫发作检测在客观质量评价指标上取得了良好的效果。

In view of the problems of too long model training time caused by the huge amount of data in the epileptic seizure detection method,and the differences between the detection results and the clinician’s marker results,an epileptic seizure detection method based on spike extraction of amplitude integrated electroencephalography(aEEG)and continuous electroencephalography(cEEG)is proposed.This proposed algorithm is developed on the basis of MATLAB software which uses morphological filtering technology to extract spikes form multi-channel aEEG to determine the seizure channel.Then the spike of the cEEG channel is extracted by morphological filtering,and the spike rate of each 5 s EEG segment is calculated.The threshold method is used for epileptic seizure detection.If the spike rate of a certain EEG segment is greater than a preset seizure threshold,the segment is the epileptic segment.The experimental results show that the algorithm achieves good results in the objective quality evaluation index.