基于神经元群模型的癫痫分析与控制
发布时间：2019-07-01 10:08
【摘要】：癫痫是一种常见的脑部疾病,具有反复性、突发性等特点。癫痫发作是由神经元的高度兴奋和高度同步化放电引起的,在脑电图中表现为高幅的不规则过度放电。因此对癫痫发病机制和治疗的研究一直是神经病学的重大难题和研究热点。而集总系数的神经元群模型可以产生与实测脑电信号类似的波形,而且其模型参数都具有一定的生理学意义,这就在信号处理方法与神经生理学研究成果之间搭建起了一座桥梁。这也为我们研究癫痫发作的生理学机制及其控制方法提供了平台。在对神经元群模型进行改进的基础上,围绕癫痫发作期和发作间歇期的生理学特征分析,癫痫兴奋性、同步性的控制方案等内容展开以下研究：1)为了分析潜在于脑电信号下的神经生理学机制,分别将脑电信号看作单个神经元群和多个神经元群的输出,进而识别癫痫发作不同时期脑电信号所对应的模型参数,通过比较模型参数的分布情况来讨论癫痫发作不同时期神经生理学机制的差异。首先,将脑电信号看作单神经元群(Wendling)模型的输出,为了仿真脑电信号测量中测量时间和设备增益对脑电信号波形的影响,在Wendling模型的基础上引入了延迟单元和增益单元。改进的Wendling模型参数的确定可以看作一个最优化问题,采用遗传算法来求解最优的模型参数组合使仿真脑电信号和实测脑电信号之间的误差最小。采用本文提出的方法确定了癫痫发作期和发作间歇期不同脑电信号所对应的模型参数。实验结果显示改进的Wendling模型可以较好地模拟实测的EEG信号,并且基于遗传算法的模型参数确定方法具有一定的稳定性。对发作期和发作间歇期脑电信号所对应的模型参数进行了比较,并讨论了癫痫发作不同时期神经元兴奋性和抑制性的差异。将脑电信号都看作一个神经元群模型的输出,而且这个神经元群中的神经元都具有统一的神经生理学参数是不合理的,会因为忽略了神经元之间的差异性而导致模型输出信号成分简单。为了更好地拟合实测的脑电信号,提出了并行连接的多神经元群模型。多神经元群模型包括多个神经元群,模型输出为各神经元群输出的线性组合。将实测的脑电信号看作多神经元群模型的输出,该模型中神经元群数并不固定,而是在保证波形足够匹配的前提下,使神经元群数最小。上述问题可以简化为一个有约束的lo范数最小化问题,采用正交匹配追踪方法来解该问题。实验结果表明,发作期的神经元群数明显少于发作间歇期,而主要神经元群的强度则较发作间歇期有大幅提升。这说明在发作过程中,会有神经元群融合的过程出现,大量相似的神经元聚集在一个神经元群导致高幅度的发放。另外,从实验结果可以看出,发作期和发作间歇期脑电数据的兴奋强度、抑制强度分布并无很大差别,但是发作期的兴奋／抑制比有一定程度的升高。2)大脑神经元的过度兴奋一直被看作引发癫痫发作的主要原因,当神经系统的自身调节能力不足以维持兴奋性-抑制性平衡的时候,就会引发癫痫发作。为了控制神经元过度兴奋引发的癫痫发作,制定了两种策略。其一为降低过度兴奋神经元的兴奋性,其二为增加抑制性以弥补神经系统自身调节的不足。提出了痫性指数来描述癫痫发作程度,并用作PID控制器的被控参数来对癫痫发作进行控制。以神经元群模型为平台仿真了兴奋性增加导致的癫痫发作程度(痫性指数)变化,进而对两种兴奋性控制策略进行了仿真。实验结果表明兴奋强度增加而保持抑制强度不变会导致痫性指数的大幅度增加,导致癫痫发作。而用PID控制器分别降低兴奋强度或增加抑制强度都可以维持兴奋-抑制平衡,并缓解癫痫的发作。3)癫痫发作总是伴有多个神经元群的超同步放电,这在脑电信号中表现为高幅的不规则过度放电。为了研究癫痫发作的超同步放电机制,并根据同步性进行癫痫发作控制,构建了包括多个具有串联关系神经元群的模型。由于同步性可以传递,讨论两个点的同步性,并不能反映整个区域的耦合结构。采用PCA算法,引入了同步族的概念,并给出了同步族强度和参与率的计算方法。在神经元群同步族的概念的基础上,提出了不进行癫痫病灶识别,而是针对癫痫发作所波及的同步族进行统一控制的癫痫同步性控制方案,打破同步族中每一对神经元群之间的耦合,从而削弱神经元群之间的同步性。对一致耦合、不一致耦合的单同步族和两同步族系统进行了实验,实验结果显示,基于PCA的方法可以很好地识别存在的同步族及每个同步族中的神经元群,针对同步族的控制策略也可以很好地控制癫痫发作。4)致痫区的准确定位是保证癫痫治疗并减少副作用的首要任务。在很多情况下,传统视觉定位方法的效果并不能让人满意。信号处理的方法可以提供大量的信息来补偿脑电信号视觉诊断的不足。致痫区定位被看作驱动方识别问题,提出了一种新的非线性互依赖性测度——加权排位互依赖性,作为驱动方的指示符,因为它可以从脑电信号提取耦合信息,特别是耦合方向信息。然后,PID控制器被用来进行癫痫发作控制。癫痫发作控制方法首先需要采用加权排位互依赖性来识别致痫区。两个单向连接在一起的神经元群模型被用来进行仿真所提出的控制方案。根据应用的不同,可以通过两个参数来调节加权排位互依赖性的灵敏度,它们各自的影响分别进行了讨论。仿真结果显示采用加权排位互依赖性来进行致痫区识别可以适应于不同类型的致痫区,可以对不同的致痫区取得98.84%的识别率。仿真同样显示PID控制器可以很好地控制神经元群之间的同步性。神经元群模型可以作为临床实验的有效补充来进行神经生理学方面的研究,具有成本低、参数调整灵活等优点,本文的研究有助于推动神经元群模型在更多领域的应用。而基于神经元群模型的癫痫兴奋性、同步性控制、病灶识别仿真可以为设计外部发作控制设备提供理论基础,可以进一步应用于临床中。
[Abstract]:Epilepsy is a common disease of the brain, which has the characteristics of renaturation, bursty and so on. The seizure is caused by the highly excited and highly synchronized discharges of the neurons, which are characterized by an irregular over-discharge of the high amplitude in the electroencephalogram. Therefore, the study of the pathogenesis and treatment of epilepsy has been a major problem and research focus of neurology. The neuron group model with the total coefficient can produce a waveform similar to that of the measured brain electrical signal, and the model parameters have a certain physiological significance, and a bridge is set up between the signal processing method and the research result of the neurophysiology. This also provides a platform for our study of the physiological mechanism of the seizure and its control. On the basis of the improvement of the neuron group model, the following studies are carried out on the physiological characteristics, the excitability and the synchronicity of the epileptic seizure and the control scheme of the synchronicity:1) In order to analyze the neurophysiological mechanism underlying the brain electrical signal, The brain electrical signal is considered as the output of a single neuron group and a plurality of neuron groups, and then the model parameters corresponding to the brain electrical signals in different periods of the epileptic seizure are identified, and the difference of the neurophysiological mechanism during the different period of the epileptic seizure is discussed by comparing the distribution of the model parameters. First, the brain electrical signal is considered as the output of a single neuron group (Wendling) model, and the delay unit and the gain unit are introduced on the basis of the Wendling model in order to simulate the influence of the measurement time and the device gain on the waveform of the brain electrical signal in the measurement of the brain electrical signal. The determination of the improved Wendling model parameter can be considered as an optimization problem, and the genetic algorithm is used to solve the optimal model parameter combination to minimize the error between the simulated brain electrical signal and the measured brain electrical signal. The model parameters corresponding to the different brain electrical signals during the onset of the seizure and the onset of the attack were determined by the method presented in this paper. The experimental results show that the modified Wendling model can well simulate the measured EEG signal, and the model parameter determination method based on the genetic algorithm has certain stability. The model parameters corresponding to the brain electrical signals during the onset and the attack period were compared, and the differences of the excitability and inhibition of the neurons during the different period of the seizure were discussed. The brain electrical signal is regarded as the output of a neuron group model, and the neuron in the neuron group has a uniform neurophysiological parameter, which can cause the model output signal component to be simple because the difference between the neurons is ignored. In order to better fit the measured brain electrical signal, a multi-neuron group model for parallel connection is proposed. The multi-neuron group model comprises a plurality of neuron groups, and the model output is a linear combination of the output of each neuron group. The measured brain electrical signal is considered as the output of the multi-neuron group model, and the number of the neurons in the model is not fixed, but the number of the neurons is minimized on the premise of ensuring that the waveform is sufficiently matched. The above problems can be simplified into a constrained lo-norm minimization problem, and the problem is solved by using the orthogonal matching tracking method. The experimental results showed that the number of the neurons in the attack period was significantly less than that of the attack, while the intensity of the main neuron group was significantly higher than that in the intermittent period. This indicates that in the course of the attack, there will be a process of fusion of the neurons, and a large number of similar neurons accumulate in a neuron group leading to high amplitude distribution. In addition, it can be seen from the experimental results that the intensity of the excitation and the intensity distribution of the EEG data during the onset and the onset of the episode are not very different, But the excitement/ suppression ratio of the onset period is somewhat increased.2) The excessive excitement of the nervous system has been seen as the main cause of the onset of the seizure, and when the self-regulation capacity of the nervous system is not sufficient to maintain the excitability-inhibitory balance, the seizure can be initiated. In order to control the seizure caused by over-excitation of the neurons, two strategies were developed. One is to reduce the excitability of the over-excited neurons, and the other is to increase the inhibition to compensate for the insufficiency of the nervous system's own regulation. The epileptic seizure degree was described by the sex index, and the controlled parameters of the PID controller were used to control the seizure. The changes of the seizure degree (seizure index) caused by the increase of excitability were simulated with the neuron group model, and the two kinds of excitability control strategies were simulated. The results show that the increase of the excitation intensity and the retention of the inhibitory intensity can lead to a significant increase in the epileptic index, resulting in a seizure. Using the PID controller to reduce the excitation intensity or the increase of the inhibition intensity, the excitation-suppression balance can be maintained, and the onset of the seizure is relieved.3) The seizure is always accompanied by a super-synchronous discharge of a plurality of neuron groups, which is represented as an irregular over-discharge of the high amplitude in the brain electrical signal. In order to study the hypersynchronous discharge mechanism of the seizure, and to control the seizure according to the synchronicity, a model including a plurality of neuron groups with a series relationship was constructed. The synchronicity of the two points is discussed and the coupling structure of the whole area cannot be reflected because the synchronism can be transmitted. By adopting the PCA algorithm, the concept of the synchronous family is introduced, and the calculation method of the synchronous family intensity and the participation rate is given. on the basis of the concept of the group of neuron groups, the invention provides an epileptic synchronization control scheme which does not carry out the identification of the epileptic focus, but also carries out a unified control on the synchronous family affected by the seizure, and breaks the coupling between each pair of neuron groups in the synchronous family, Thereby weakening the synchronicity between the neuronal populations. The results show that the PCA-based method can well identify the existing synchronous family and the neuron group in each of the synchronous families. The control strategy for the synchronous family can also control the seizure.4) The accurate location of the epilepsy area is the primary task of ensuring the treatment of the epilepsy and reducing the side effect. In many cases, the effect of the traditional vision positioning method is not satisfactory. The signal processing method can provide a large amount of information to compensate for the deficiency of the visual diagnosis of the brain electrical signal. The location of the epilepsy area is considered as the problem of the identification of the driver, and a new non-linear cross-dependency measure _ weighted rank interdependency is proposed as the indicator of the driver because it can extract the coupling information from the brain electrical signal, in particular the coupling direction information. The PID controller is then used to carry out the seizure control. The epileptic seizure control method first needs to use the weighted rank interdependency to identify the epilepsy area. The two unidirectionally connected neuron group models are used to simulate the proposed control scheme. Depending on the application, the sensitivity of the weighted rank mutual dependence can be adjusted by two parameters, and their respective influences are discussed respectively. The results of the simulation show that the identification of the epilepsy can be adapted to different types of epilepsy, and the recognition rate of 98.84% can be obtained for different epilepsy areas. The simulation also shows that the PID controller can well control the synchronization between the neuron groups. The neuron group model can be used as an effective complement of clinical experiments to study the neurophysiology, and has the advantages of low cost, flexible parameter adjustment, and the like, and the research of the invention can help to promote the application of the neuron group model in the more field. In addition, on the basis of the neuron group model, the excitability, the synchronicity control and the focus recognition simulation can provide the theoretical basis for the design of the external attack control equipment, and can be further applied to the clinical application.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：R742.1
，
本文编号：2508388
[Abstract]:Epilepsy is a common disease of the brain, which has the characteristics of renaturation, bursty and so on. The seizure is caused by the highly excited and highly synchronized discharges of the neurons, which are characterized by an irregular over-discharge of the high amplitude in the electroencephalogram. Therefore, the study of the pathogenesis and treatment of epilepsy has been a major problem and research focus of neurology. The neuron group model with the total coefficient can produce a waveform similar to that of the measured brain electrical signal, and the model parameters have a certain physiological significance, and a bridge is set up between the signal processing method and the research result of the neurophysiology. This also provides a platform for our study of the physiological mechanism of the seizure and its control. On the basis of the improvement of the neuron group model, the following studies are carried out on the physiological characteristics, the excitability and the synchronicity of the epileptic seizure and the control scheme of the synchronicity:1) In order to analyze the neurophysiological mechanism underlying the brain electrical signal, The brain electrical signal is considered as the output of a single neuron group and a plurality of neuron groups, and then the model parameters corresponding to the brain electrical signals in different periods of the epileptic seizure are identified, and the difference of the neurophysiological mechanism during the different period of the epileptic seizure is discussed by comparing the distribution of the model parameters. First, the brain electrical signal is considered as the output of a single neuron group (Wendling) model, and the delay unit and the gain unit are introduced on the basis of the Wendling model in order to simulate the influence of the measurement time and the device gain on the waveform of the brain electrical signal in the measurement of the brain electrical signal. The determination of the improved Wendling model parameter can be considered as an optimization problem, and the genetic algorithm is used to solve the optimal model parameter combination to minimize the error between the simulated brain electrical signal and the measured brain electrical signal. The model parameters corresponding to the different brain electrical signals during the onset of the seizure and the onset of the attack were determined by the method presented in this paper. The experimental results show that the modified Wendling model can well simulate the measured EEG signal, and the model parameter determination method based on the genetic algorithm has certain stability. The model parameters corresponding to the brain electrical signals during the onset and the attack period were compared, and the differences of the excitability and inhibition of the neurons during the different period of the seizure were discussed. The brain electrical signal is regarded as the output of a neuron group model, and the neuron in the neuron group has a uniform neurophysiological parameter, which can cause the model output signal component to be simple because the difference between the neurons is ignored. In order to better fit the measured brain electrical signal, a multi-neuron group model for parallel connection is proposed. The multi-neuron group model comprises a plurality of neuron groups, and the model output is a linear combination of the output of each neuron group. The measured brain electrical signal is considered as the output of the multi-neuron group model, and the number of the neurons in the model is not fixed, but the number of the neurons is minimized on the premise of ensuring that the waveform is sufficiently matched. The above problems can be simplified into a constrained lo-norm minimization problem, and the problem is solved by using the orthogonal matching tracking method. The experimental results showed that the number of the neurons in the attack period was significantly less than that of the attack, while the intensity of the main neuron group was significantly higher than that in the intermittent period. This indicates that in the course of the attack, there will be a process of fusion of the neurons, and a large number of similar neurons accumulate in a neuron group leading to high amplitude distribution. In addition, it can be seen from the experimental results that the intensity of the excitation and the intensity distribution of the EEG data during the onset and the onset of the episode are not very different, But the excitement/ suppression ratio of the onset period is somewhat increased.2) The excessive excitement of the nervous system has been seen as the main cause of the onset of the seizure, and when the self-regulation capacity of the nervous system is not sufficient to maintain the excitability-inhibitory balance, the seizure can be initiated. In order to control the seizure caused by over-excitation of the neurons, two strategies were developed. One is to reduce the excitability of the over-excited neurons, and the other is to increase the inhibition to compensate for the insufficiency of the nervous system's own regulation. The epileptic seizure degree was described by the sex index, and the controlled parameters of the PID controller were used to control the seizure. The changes of the seizure degree (seizure index) caused by the increase of excitability were simulated with the neuron group model, and the two kinds of excitability control strategies were simulated. The results show that the increase of the excitation intensity and the retention of the inhibitory intensity can lead to a significant increase in the epileptic index, resulting in a seizure. Using the PID controller to reduce the excitation intensity or the increase of the inhibition intensity, the excitation-suppression balance can be maintained, and the onset of the seizure is relieved.3) The seizure is always accompanied by a super-synchronous discharge of a plurality of neuron groups, which is represented as an irregular over-discharge of the high amplitude in the brain electrical signal. In order to study the hypersynchronous discharge mechanism of the seizure, and to control the seizure according to the synchronicity, a model including a plurality of neuron groups with a series relationship was constructed. The synchronicity of the two points is discussed and the coupling structure of the whole area cannot be reflected because the synchronism can be transmitted. By adopting the PCA algorithm, the concept of the synchronous family is introduced, and the calculation method of the synchronous family intensity and the participation rate is given. on the basis of the concept of the group of neuron groups, the invention provides an epileptic synchronization control scheme which does not carry out the identification of the epileptic focus, but also carries out a unified control on the synchronous family affected by the seizure, and breaks the coupling between each pair of neuron groups in the synchronous family, Thereby weakening the synchronicity between the neuronal populations. The results show that the PCA-based method can well identify the existing synchronous family and the neuron group in each of the synchronous families. The control strategy for the synchronous family can also control the seizure.4) The accurate location of the epilepsy area is the primary task of ensuring the treatment of the epilepsy and reducing the side effect. In many cases, the effect of the traditional vision positioning method is not satisfactory. The signal processing method can provide a large amount of information to compensate for the deficiency of the visual diagnosis of the brain electrical signal. The location of the epilepsy area is considered as the problem of the identification of the driver, and a new non-linear cross-dependency measure _ weighted rank interdependency is proposed as the indicator of the driver because it can extract the coupling information from the brain electrical signal, in particular the coupling direction information. The PID controller is then used to carry out the seizure control. The epileptic seizure control method first needs to use the weighted rank interdependency to identify the epilepsy area. The two unidirectionally connected neuron group models are used to simulate the proposed control scheme. Depending on the application, the sensitivity of the weighted rank mutual dependence can be adjusted by two parameters, and their respective influences are discussed respectively. The results of the simulation show that the identification of the epilepsy can be adapted to different types of epilepsy, and the recognition rate of 98.84% can be obtained for different epilepsy areas. The simulation also shows that the PID controller can well control the synchronization between the neuron groups. The neuron group model can be used as an effective complement of clinical experiments to study the neurophysiology, and has the advantages of low cost, flexible parameter adjustment, and the like, and the research of the invention can help to promote the application of the neuron group model in the more field. In addition, on the basis of the neuron group model, the excitability, the synchronicity control and the focus recognition simulation can provide the theoretical basis for the design of the external attack control equipment, and can be further applied to the clinical application.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：R742.1
，
本文编号：2508388
本文链接：https://www.wllwen.com/yixuelunwen/shenjingyixue/2508388.html
最近更新
教材专著
