EEG-Based Detection of Epileptic Seizures Through the Use of a Directed Transfer Function Method
EEG-Based Detection of Epileptic Seizures Through the Use of a Directed Transfer Function Method
复制标题
DOI：
10.1109/access.2018.2867008
复制
发表时间：
2018
期刊：
影响因子：
3.9
通讯作者：
G. Wang;Doutian Ren;Kuo Li;Dong Wang;Maode Wang;Xiangguo Yan
中科院分区：
文献类型：
--
作者：
G. Wang;Doutian Ren;Kuo Li;Dong Wang;Maode Wang;Xiangguo Yan
本文旨在探索癫痫发作的自动检测方法，以提高难治性癫痫患者的治疗和诊断水平。提出了一种基于直接传递函数(DTF)的癫痫发作检测新算法。首先，采用滑动窗口技术对脑电记录进行分割，并利用DTF算法计算脑功能连通性。然后，通过将单个脑电通道到其他通道的信息流相加，计算基于DTF得到的连通性的总信息流。最后，将信息流出作为支持向量机分类器的特征，用于区分发作期和发作期的脑电片段。对10例癫痫患者的平均正确率为98.45%，平均选择性为64.43%，平均敏感度为93.36%，平均特异度为98.42%，平均检测率为95.89%。通过统计分析，基于DTF的方法在五个评价标准上与其他算法相比具有统计学意义上的优势。我们的结果表明，DTF衍生的连通性可以表征癫痫发作状态下脑区之间的动态因果相互作用模式，该方法适用于癫痫发作的检测。
This paper aims to explore the automatic detection method of epileptic seizures to improve the treatment and diagnosis of medically refractory epilepsy patients. A new algorithm based on directed transfer function (DTF) method was proposed for epileptic seizure detection. First, the sliding window technique was used to segment electroencephalogram (EEG) recordings, and the cerebral functional connectivity was calculated by the DTF algorithm. Then, the total information outflow based on the DTF-derived connectivity was calculated by adding up the information flow from a single EEG channel to other channels. Finally, the information outflow was assigned as the features of support vector machine (SVM) classifier to discriminate interictal and ictal EEG segments. For 10 epilepsy patients, the proposed algorithm provided the mean correct rate of 98.45%, the mean selectivity of 64.43%, the mean sensitivity of 93.36%, the mean specificity of 98.42%, and the average detection rate of 95.89%. By applying the statistical analysis, the superiority of DTF-based method was statistically significant when compared with other algorithms in terms of five assessment criteria. Our results indicated that the DTF-derived connectivity could characterize the dynamic causal interaction patterns between brain areas during seizure states, and the proposed method was suitable for the detection of epileptic seizures.