PSYCH OpenIR  > 健康与遗传心理学研究室
激光诱发电位的特征提取与疼痛预测—基于多种脑电分析方法的跨物种研究
其他题名Feature extraction and pain prediction of laser-evoked potentials: a study based on various methods of EEG signal processing in humans and rats
夏晓磊
导师胡理
2023-06
摘要慢漫性疼痛已经成为一个全社会普遍关注的公共健康问题。疼痛不仅给人们带来身体上的不适,而且对人们的精神、心理等诸多方面产生不同程度的负面影响,严重降低了人们的生活质量。对疼痛原理和镇痛策略的研究将有助于解决现实生活中日益突出的疼痛问题和社会问题。然而,现阶段对疼痛的诊断和治疗在很大程度上依靠患者的主观报告,其主观性给疼痛的客观评估和科学诊断带来了很大的困难。因此,开发客观可靠的疼痛测量技术以此来有效预测疼痛知觉的研究意义重大,目前,利用多种神经成像技术来揭示潜在的疼痛知觉的特异性指标已经逐渐成为国内外疼痛研究的热点之一。 激光可以选择性的激活和疼痛相关的As纤维和c纤维,避免激活和触觉相关的A}纤维,因此,激光诱发电位(Laser-evoked potentials, LEPs)是一项非常重要的疼痛研究技术,被认为是疼痛研究中的黄金标准。该技术不仅被广泛应用于以人类为被试的研究中,同样广泛应用于以大鼠被试为代表的动物研究中。然而,我们前期的一项研究已经证明,过去几十年来很多大鼠LEPs研究中发现的所谓的“A s -LEPs",实际上是由于激光引发的超声波引起的大鼠听觉脑响应而非疼痛脑响应,由于听觉敏感性的差异,这种超声波人类很难察觉但是可以被大鼠捕捉到。因此,之前的很多关于大鼠"A s -LEPs”的研究都是有问题的,这些研究中所探讨的大鼠疼痛脑响应的一些特性也是错误的。我们前期的研究进一步提出了优化的动物急性疼痛模型,即在白噪声掩蔽条件下对自由活动的大鼠使用激光诱发电位技术,该模型被证明是有效的可以促进从动物到人的疼痛转化研究的效率。 本研究是在我们之前研究基础上的进一步拓展深入,即在使用优化的大鼠急性疼痛模型的前提下,我们利用多种脑电信号处理技术分别提取人类被试和大鼠被试的多种激光诱发电位特征,并利用支持向量机等机器学习技术对提取到的疼痛特征进行训练和疼痛强度预测。研究一是采用传统的脑电信号处理技术从时域和时频域两个维度对58名人类被试的脑电信号进行特征提取,并将提取到的特征用于支持向量机算法模型的训练并预测疼痛的强度,结果表明,时域提取到的A s -ERPs响应以及时频域提取到的A s -ERPs对应的低频时频表征和高频γ振荡等特征都可以很好的预测疼痛强度。研究二是采用传统的脑电信号处理技术从时域和时频域两个维度对12只大鼠被试的脑电信号进行特征提取,并将提取到的特征用于支持向量机算法模型的训练并预测疼痛的强度,结果表明,时域提取到的C-ERPs响应以及时频域提取到的C-ERPs对应的低频时频表征和高频γ振荡等特征都可以很好的预测疼痛强度。研究三是采用复杂度分析和熵等多种非线性动力学的数据分析方法,对58名人类被试的脑电信号进行多种非线性特征提取,并将提取到的特征用于支持向量机算法模型的训练并预测疼痛的强度,结果表明,提取到的部分非线性动力学特征可以很好的预测疼痛强度,并且这些非线性动力学特征与人类被试时域上A s -ERPs和C-ERPs所对应的时间窗口和特性相一致。研究四是采用复杂度分析和嫡等多种非线性动力学的数据分析方法,对12只大鼠被试的脑电信号进行多种非线性特征提取,并将提取到的特征用于支持向量机算法模型的训练并预测疼痛的强度,结果表明,提取到的部分非线性动力学特征同样可以很好的预测疼痛强度。研究五的结果表明,同时采用多种脑电分析方法的多个维度的特征选择,可在一定程度上提高疼痛预测的准确率。 综上所述,本研究利用从传统脑电分析方法和非线性动力学分析方法获取到的多种不同维度的疼痛特征,系统考察了激光诱发电位技术在人类研究和大鼠研究中的疼痛特征和预测效果。其中,非线性动力学特征的提取和预测,在该研究领域还是首次,某些非线性动力学提取的特征在预测疼痛强度上具有良好的效果,这提示我们在激光诱发电位的研究领域中要关注那些如非线性动力学特征等的非传统的脑电分析方法。另外,和人类研究相似的良好的大鼠疼痛特征和疼痛预测结果进一步表明,我们之前提出的优化的动物急性疼痛模型是有效的。
其他摘要Pain, especially chronic pain, has become a public health problem that has been concerned by the whole society. It not only lead to somatic discomfort, but also adversely affect the health of humans in psychological, physical, and many other aspects, thus directly reducing the people's quality of life. Research on the mechanism of pain and analgesic strategy would not only help relieving patients from pain, but also help solving the serious, pain induced, social issues. However, at the current stage, the diagnosis and treatment of pain heavily rely on the subjective and inaccurate report of pain from patients. Since an objective and reliable measurement of pain is highly needed in various basic and clinical applications, several neuroimaging techniques have been adopted to reveal neurologic signature of pain perception and to predict pain based on nociceptive-evoked brain responses. This is becoming one of the worldwide hot topics of pain research. Laser stimulators currently represent the most accurate tool for experimental studies of pain,6 as they activate Ad- and C nociceptors selectively and elicit sensations of "pure" pain without touch. For these reasons, laser stimulation is considered the gold standard to investigate pain psychophysically and electrophysiologically, and it has been used in hundreds of human and rat studies. However, our previous study demonstrated that the so-called "Aδ-LEPs", instead of reflecting the activation of the Aδ-nociceptive system, is actually consequent to the activation of the auditory system by laser-generated ultrasounds that can be detected by rats, but not by humans. This auditory response has been so far mistakenly interpreted as reflecting the A8-somatosensory input, thus undermining the conclusions of several previous investigations. As demonstrated in our previous study, recording LEPs in freely-moving rats with white noise masking is a valid model to improve the translation of animal results to human physiology and pathophysiology. By using the optimized animal models, several further studies are performed in this paper. By using various methods of EEG signal processing, diverse features of LEPs are extracted from both humans and rats in different ways, and are trained in a support vector machine(SVM) model to predict the intensity of pain. Specifically, traditional EEG signal processing methods containing time domain analysis and time frequency analysis are performed to extract features from 58 human subjects in study 1 .The results shows that human A s -ERP responses in time domain and A δ –ERP responses and y oscillation in time frequency domain play an important role in predicting pain intensity. Furthermore, traditional EEG signal processing methods containing time domain analysis and time frequency analysis are performed to extract features from 12 rats in study 2. The result shows that rat C-ERP responses in time domain and C-ERP responses in time frequency domain and y oscillations in time frequency domain play an important role in predicting pain intensity. In study 3, several methods of complexity analysis and entropy analysis are used to extract features from 58 human subjects, and some of them are important in predicting pain intensity. In study 4, some of features extracted from 12 rats by using diverse methods of complexity analysis and entropy analysis, also make great contributions to predicting pain intensity in rats. In study 5, the result shows that the use of multi-dimension features can improve the prediction accuracy of pain in both humans and rats. In summary, by using various methods of EEG signal processing, this study systematically investigated diverse features of LEPs extracted from both humans and rats, and the role of these features in predicting the intensity of pain. Several methods of complexity analysis and entropy analysis are provided to be important in predicting pain intensity, and they should be more widely used in future studies.
关键词激光诱发电位 特征提取 疼痛预测 动物模型
学位类型博士
语种中文
学位名称理学博士
学位专业认知神经科学
学位授予单位中国科学院大学
学位授予地点中国科学院心理研究所
文献类型学位论文
条目标识符http://ir.psych.ac.cn/handle/311026/46213
专题健康与遗传心理学研究室
推荐引用方式
GB/T 7714
夏晓磊. 激光诱发电位的特征提取与疼痛预测—基于多种脑电分析方法的跨物种研究[D]. 中国科学院心理研究所. 中国科学院大学,2023.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
Enhancing Detailed F(54461KB)学位论文 限制开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[夏晓磊]的文章
百度学术
百度学术中相似的文章
[夏晓磊]的文章
必应学术
必应学术中相似的文章
[夏晓磊]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。