癫痫大脑神经电活动的定位和预测方法研究

Epilepsy is one of the most common and devastating neurological disorders, afflicting more than 50 million individuals worldwide. For epilepsy patients who do not achieve complete seizure control, the seizures often lead to threaten life, especially if they occur while the person is driving, swimming, climbing heights, or alone.Apart from the risk of serious injury, there is often an intense feeling of helpness that has a strong impact on the everyday life of a patient. Thereby, a method capable of removing or controlling the occurrence of seizures could significantly improve the quality of life for epilepsy patients. Nowadays, majority of epilepsy patients could achieve sufficient seizure control from anticonvulsive medication and resective surgery, but there are still some limitations with antiepileptic drugs treatment and surgical therapy. Thus, sufficient treatment is not currently available for a lot of patients with medically intractable epilepsy. This project will try to study the method of expolring brain electrical activity around seizure onset in two aspects, aiming at clinical epilepsy therapy. On the one hand, brain electrical activity around seizure onset will be explored in the spatial domain, i.e. EEG (electroencephalogram) source localization of epileptogenic zone. A new algorithm of EEG source localization based on wavelet packet transform will be proposed. Then the epileptic source will be stimulated by using dipole model and the validity of this EEG source localization algorithm will be evaluated in this simulation study. In the clinical experiments, the epielptogenic zone will be identified from the pre-operative and post-operative MR images in a cohort of epilepsy patients and the ability of source localization of this algorithm will be examined from EEG recording. On the other hand, the brain electrical activity of preictal state preceding a seizure onset will be accurately detected in the time domain, i.e. seizure prediction. In this research, seizure prediction will be analyzed in terms of the mechanisms of ictogenesis that take into consideration the complex spatiotemporal interactions between different brain regions in epilepsy. The methods will include neural signal synchronization anlaysis and neural information flow pattern analysis which reflect the nature of brain electrical activity in epilepsy. These methods can quantify relations between recording sites to characterize interaction among different brain regions. Furthermore, validity and reliability of these methods will be verified in clinical experiments.It is expected that the research results of this project can provide new ideas and theoretical foundations for both effective treatment and cure of epilepsy. In the end, the principle investigator has already had some experiences about EEG source localization using interictal spike in patients with partial epilepsy and obtained preliminary simulation results by using the new EEG source localization algorithm.

癫痫是一种常见的神经疾病,目前常见的药物治疗和手术治疗方法都存在一定的现实问题，使众多癫痫患者仍无法得到有效治疗。本项目针对癫痫的临床治疗需求，从两方面研究了癫痫大脑活动的探测方法。一方面从空间域上对癫痫活动进行准确地探测，即癫痫病灶的源定位，本项目提出了一种新的基于小波包变换的脑电源定位算法，并通过仿真研究和临床实验来验证该癫痫病灶源定位方法在临床实际使用中的有效性。另一方面从时间域上对癫痫活动进行准确的早期探测，即癫痫发作的预测，本项目拟结合大脑神经信号同步分析和大脑神经信息流分析，在量化不同脑区间关系以刻画其相互作用的基础上，进行癫痫发作预测的研究，并依靠临床实验研究来证明其能够提高癫痫预测的有效性及可靠性。预计该课题的研究成果能够为癫痫患者的有效治疗和治愈提供新的思路和理论基础。项目申请人已具有利用癫痫发作间期棘波进行脑电源定位的研究经历，并已获得新的源定位算法的初步仿真研究结果。

癫痫是一种常见的神经疾病,目前常见的药物治疗和手术治疗方法都存在一定的现实问题，使众多癫痫患者仍无法得到有效治疗。本项目针对癫痫的临床治疗需求，研究了癫痫大脑活动的探测方法。研究获得以下成果：1）针对癫痫患者脑电数据伪迹干扰严重的问题，研究了脑电信号中肌电和眼电伪迹去除的新方法，提出了基于多元经验模态分解的肌电伪迹去除算法，实验结果表明，不同信噪比下通过该方法去噪后对每个脑电通道信噪比有明显地提升以及均方误差显著地下降，另一方面，提出了基于独立分量分析和多元经验模态分解的眼电伪迹去除算法，该方法能够完全去除掉脑电信号中的眼电伪迹，眼电伪迹去除后脑电与眼电之间的互相关系数显著性的减小；2）研究了新的基于小波包变换的脑电源定位算法，通过对脑电信号的小波包分解、子空间分量选择、信号重构、真实头部边界元模型建立以及脑电逆问题求解，由头皮脑电精确反演大脑皮层上的神经电活动，仿真和视觉诱发电位源定位的研究表明该方法具有更高的源定位精度，并通过临床实验研究来验证了该方法在临床实际使用中的优越性，从而能够为癫痫的外科手术治疗提供重要的临床诊断信息；3）研究了一种基于脑电活动同步信息的癫痫预测算法，通过双变量经验模态分解及希尔伯特变换获取脑电记录的瞬时相位信息，然后依据提取的相位信息计算平均相位相干性是否超出某一预先设定的阈值进行癫痫发作的预测，并依靠临床实验研究来证明了其能够提高癫痫预测的有效性及可靠性，为癫痫患者及时规避风险，得以控制癫痫发作提供可能性；4）研究了基于脑功能连接分析的癫痫发作检测分析方法，该方法利用部分有向相干分析来提取癫痫脑电信号的特征，之后根据癫痫发作的病理学特征，以流出信息作为支持向量机分类器的输入，对癫痫发作间期和发作期的脑电信号进行模式识别分类，结果表明该方法在正确率、敏感性和特异性等方面有明显的优越性，该癫痫发作自动检测方法可以有效地减少医生阅读脑电图的工作量，提高癫痫发作判读的客观性。这些成果为进一步研究癫痫疾病的有效治疗提供了理论依据和实践基础。在项目执行期间，在国内外重要学术期刊上和重要国际学术会议上发表论文11篇，其中：国外学术期刊论文8篇，国内核心学术期刊论文1篇，国际学术会议分组论文2篇，其中SCI检索8篇，EI检索3篇。申请中国发明专利5项，已获批准授权2项，授权软件著作权1项。培养硕士研究生5名，其中4名已毕业。