基于发作间期EEG的癫痫自动诊断系统的研究与设计

发布时间：2018-06-04 23:04

本文选题：发作间期脑电图 + 癫痫　；参考：《南京邮电大学》2016年硕士论文

【摘要】：脑电图是临床医生对大脑神经系统损伤及病变进行检查和诊断的有效工具,其中包含了丰富的病理和生理信息。因为癫痫是由于神经元异常放电活动导致的神经系统疾病,所以脑电图检查是临床上诊断癫痫的一种基本方法。传统的癫痫诊断依赖于训练有素的医生观察冗长的脑电图记录,识别脑电图中的癫痫相关异常波形来诊断病患,这种方法枯燥而且耗时。虽然如今有了许多的自动癫痫检测系统辅助医生进行癫痫诊断,但是这些系统依然要求记录长程的脑电图以获取癫痫发作数据。由于长程脑电图在医疗水平欠发达的地方难以实现,所以对癫痫脑电信号进行特征提取分析,设计自动癫痫诊断系统将具有非常重要的意义,尤其是如果能够仅仅依赖于发作间期的脑电图信号来实现癫痫诊断。本文研究脑电图的特征提取,并且设计了基于概率神经网络的分类器实现自动癫痫诊断系统。自动癫痫诊断系统分为特征提取和分类器模块。在对脑电信号处理的主要研究方法进行总结之后,本文采用基于功率谱分析的传统频谱特征和基于时间序列分析的动态特征构造脑电图的24维的特征向量,作为分类器的输入。对系统进行测试的实验结果表明,该系统能够有效的区分健康人的脑电图片段、癫痫病人发作间期的脑电图片段以及癫痫病人发作期的脑电图片段,也能够为癫痫病灶的定位提供一定的参考信息。另外,本文从特征参数的选择和散布常数的选取两方面讨论了该系统的优化问题。文本还使用了伪差和噪声更多却更贴近实际应用的头皮EEG来检测自动癫痫诊断系统。通过将概率神经网络和一种投票机制相结合,系统在区分发作间期头皮脑电图和健康人脑电图时能够达到99.38%的准确率。这个结果证明了使用癫痫发作间期头皮脑电图进行癫痫诊断的可行性。
[Abstract]:Electroencephalogram (EEG) is an effective tool for clinicians to examine and diagnose the injury and pathological changes of the cerebral nervous system, which contains abundant pathological and physiological information. Because epilepsy is a nervous system disease caused by abnormal discharges of neurons, EEG is a basic method for clinical diagnosis of epilepsy. Traditional diagnosis of epilepsy relies on trained doctors to observe long EEG records and identify epileptic-related abnormal waveforms in EEG to diagnose patients. This method is tedious and time-consuming. Although there are a number of automated epilepsy detection systems to assist doctors in the diagnosis of epilepsy, these systems still require long-term EEG records to obtain seizure data. Because it is difficult to realize long range EEG in places where medical level is not developed, it is very important to extract and analyze the characteristics of epileptic EEG signals and design an automatic epilepsy diagnosis system. Especially if you can only rely on the interictal EEG signal to achieve epilepsy diagnosis. In this paper, the feature extraction of EEG is studied, and a classifier based on probabilistic neural network is designed to realize the automatic epilepsy diagnosis system. Automatic epilepsy diagnosis system is divided into feature extraction and classifier module. After summarizing the main research methods of EEG processing, this paper uses the traditional spectrum features based on power spectrum analysis and the dynamic features based on time series analysis to construct the 24-dimensional eigenvector of EEG as the input of the classifier. The experimental results of the system showed that the system could effectively distinguish the EEG fragments of healthy people from those of epileptic patients during interictal period and those of epileptic patients during seizures. It can also provide some reference information for the location of epileptic foci. In addition, this paper discusses the optimization of the system from two aspects: the selection of characteristic parameters and the selection of dispersion constants. The paper also uses scalp EEG, which is more false and noise, to detect automatic epilepsy diagnosis system. By combining probabilistic neural network with a voting mechanism, the system can achieve 99.38% accuracy in distinguishing scalp electroencephalogram (EEG) between interictal period and healthy person. The results demonstrate the feasibility of using interictal scalp electroencephalogram (EEG) for the diagnosis of epilepsy.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH776

【相似文献】