癫痫发作脑电信号的相位幅值耦合特征的研究

发布时间：2018-01-01 03:33

本文关键词：癫痫发作脑电信号的相位幅值耦合特征的研究　出处：《沈阳工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：癫痫是大脑神经元突发性异常放电,导致短暂的大脑功能性障碍的一种慢性疾病。癫痫发作具有突然性和反复性,给患者学习、生活和工作形成极大障碍,严重影响到患者及其家庭的生活质量。由于其发作短暂,临床医生很少目睹发作过程,增加了诊治的困难。脑电图可记录患者癫痫发作异常放电过程,已成为癫痫诊治最重要的工具。然而,长时程视频脑电图通常记录数天乃至数周,给癫痫中心带来了海量数据,癫痫发作数据的标记与分析成为了临床医生一项繁重工作。本论文以癫痫发作脑电图为研究对象,分析了癫痫发作期脑电图的低频相位与高频幅度耦合的特征,研究了不同低频和高频节律间的耦合关系;基于相位-幅度耦合特征对癫痫发作间期与发作期脑电图进行了分类研究,实现了发作期数据段的自动标记。具体工作包括:首先,本文基于波恩脑电数据集,研究了癫痫脑电低频节律相位与高频节律幅度间的耦合关系,利用调制指数(Modulation Index,MI)来量化各频段间的耦合强度,提出了依据高低频节律范围对MI图分区方法。分析结果显示,与发作间期相比,发作期Gamma节律与多种低频节律的MI值均显著(p0.01)增强。Theta节律与Beta节律间MI值存在显著差异,分区后MI特征对发作期和间期数据分类正确率达到97%。其次,本文从脑电图非线性特征出发,研究了癫痫发作期脑电图的排列熵特征,将排列熵、标准差等特征联合,对波恩脑电数据集癫痫发作间期和发作期数据进行特征提取。结果显示在分类过程中癫痫脑电图的排列熵和标准差特征具有互补性,在计算排列熵符号化过程中,有尺度信息损失,而标准差等特征可弥补相关信息,二者联合也可使癫痫发作与癫间脑电识别率达到97%。再次,本文对癫痫狗脑电图进行了相位-幅值耦合特征的分析。利用大样本数癫痫狗数据分类结果显示,耦合特征能够用于脑电图自动分类,正确率达到92%,且多个低频和高频节律间耦合特征对分类结果都有影响。最后,本文将癫痫发作期脑电图的相位-幅度耦合、排列熵等特征应用于实测数据分析,对沈阳军区总医院患者术前评估记录的颅内脑电图进行了发作期自动标记研究;在对单次发作数据标记的基础上,可以非常准确的自动标记出患者其他的发作,分类准确率为95.5%。本文研究方法在临床的应用将能有效降低医生长时程脑电图分析负担,具有很好应用前景和现实意义。
[Abstract]:Epilepsy is a chronic disease caused by sudden abnormal discharges of brain neurons, which leads to transient functional disorders of the brain. Epileptic seizures have sudden and repetitive characteristics, which make the patients learning, living and working severely impaired. It seriously affects the quality of life of patients and their families. Because of its short duration, clinicians rarely see the onset process, increasing the difficulty of diagnosis and treatment. EEG can record the abnormal discharge process of epileptic seizures in patients. It has become the most important tool for the diagnosis and treatment of epilepsy. However, long time video EEG usually records several days or even weeks, which brings a lot of data to the epileptic center. The marking and analysis of epileptic seizure data has become a heavy task for clinicians. In this paper, the characteristics of low frequency phase and high frequency amplitude coupling of EEG during epileptic seizures were analyzed. The coupling relationship between different low-frequency and high-frequency rhythms is studied. Based on the phase-amplitude coupling feature, the electroencephalogram (EEG) in the interictal phase and the seizure phase is classified, and the automatic marking of the seizure data segment is realized. The specific work includes: first, based on the Bonn EEG data set. The coupling relationship between the phase of low-frequency rhythm and the amplitude of high-frequency rhythm in epileptic EEG was studied. The modulation index Modulation Index (MI) was used to quantify the coupling intensity between different frequency bands. A method of dividing MI map according to the range of high and low frequency rhythm is proposed. The results show that it is compared with interictal period. The MI values of Gamma rhythm and various low frequency rhythms were significantly increased (p 0.01). There was significant difference between the MI value of Theta rhythm and Beta rhythm. The classification accuracy rate of MI features on the seizure and interphase data is 97%. Secondly, from the nonlinear characteristics of EEG, this paper studies the permutation entropy characteristics of EEG during epileptic seizures, which will be permutation entropy. The standard deviation and other features were combined to extract the interictal and interictal data from the Bonn EEG data set. The results showed that the entropy and standard deviation of EEG were complementary in the course of classification. In the process of computing the entropy symbolization, there is a loss of scale information, and the standard deviation can compensate for the relevant information. The combination of the two can also make the recognition rate of epileptic seizures and epileptic diencephalogram reach 97%. Thirdly. In this paper, the phase amplitude coupling characteristics of EEG in epileptic dogs were analyzed. The results showed that the coupling features could be used to classify EEG automatically, and the correct rate was 92%. And the coupling characteristics between low frequency and high frequency rhythm have influence on the classification results. Finally, the phase amplitude coupling and permutation entropy of EEG during epileptic seizures are applied to the analysis of measured data. The intracranial electroencephalogram (EEG) recorded by preoperative evaluation in Shenyang military region General Hospital (Shenyang military region General Hospital) was studied with automatic marking during the attack period. On the basis of the single attack data marker, it is very accurate to automatically mark the other attacks of the patient. The classification accuracy is 95.5.The clinical application of this research method can effectively reduce the burden of long-term EEG analysis of doctors, which has a good application prospect and practical significance.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R742.1

【参考文献】