College of Electronic Information Engineering, Changchun University of Science and Technology, Changchun 130022, China.
College of Physics and Electronic Information, Baicheng Normal University, Baicheng 137099, China.
Sensors (Basel). 2023 Sep 25;23(19):8078. doi: 10.3390/s23198078.
Epilepsy does great harm to the human body, and even threatens human life when it is serious. Therefore, research focused on the diagnosis and treatment of epilepsy holds paramount clinical significance. In this paper, we utilized variational modal decomposition (VMD) and an enhanced grey wolf algorithm to detect epileptic electroencephalogram (EEG) signals. Data were extracted from each patient's preseizure period and seizure period of 200 s each, with every 2 s as a segment, meaning 100 data points could be obtained for each patient's health period as well as 100 data points for each patient's epilepsy period. Variational modal decomposition (VMD) was used to obtain the corresponding intrinsic modal function (VMF) of the data. Then, the differential entropy (DE) and high frequency detection (HFD) of each VMF were extracted as features. The improved grey wolf algorithm is adopted for a selected channel to improve the maximum value of the channel. Finally, the EEG signal samples were classified using a support vector machine (SVM) classifier to achieve the accurate detection of epilepsy EEG signals. Experimental results show that the accuracy, sensitivity and specificity of the proposed method can reach 98.3%, 98.9% and 98.5%, respectively. The proposed algorithm in this paper can be used as an index to detect epileptic seizures and has certain guiding significance for the early diagnosis and effective treatment of epileptic patients.
癫痫对人体危害极大，严重时甚至威胁生命。因此，研究癫痫的诊断和治疗具有重要的临床意义。本文采用变分模态分解（VMD）和改进灰狼算法对癫痫脑电信号进行检测。从每位患者的发作前和发作期各提取 200 s 的数据，每 2 s 为一段，即每位患者健康期可获得 100 个数据点，每位患者癫痫期可获得 100 个数据点。采用变分模态分解（VMD）获取数据的对应固有模态函数（VMF）。然后提取每个 VMF 的差分熵（DE）和高频检测（HFD）作为特征。采用改进的灰狼算法对选定通道进行优化，提高通道的最大值。最后，采用支持向量机（SVM）分类器对 EEG 信号样本进行分类，实现对癫痫脑电信号的准确检测。实验结果表明，该方法的准确率、灵敏度和特异性分别可达 98.3%、98.9%和 98.5%。本文提出的算法可以作为癫痫发作的检测指标，对癫痫患者的早期诊断和有效治疗具有一定的指导意义。
