基于静息状态fMRI的ADHD复杂脑网络分析

发布时间：2018-04-25 05:09

本文选题：静息状态功能磁共振成像 + 小世界属性　；参考：《北京交通大学》2015年硕士论文

【摘要】：注意缺陷／多动障碍(Attention Deficit/Hyperactivity Disorder, ADHD)是一种很常见的儿童精神发育障碍疾病。近十几年来,ADHD已经是我国乃至全球的严重的公共卫生问题。近年来,基于图论的复杂网络分析极大地推进了人类对于脑网络组织模式的认识。本研究中,我们基于96个儿童的静息状态fMRI图像,采用图论的方法构建了每个儿童的复杂脑网络,进而分析了ADHD症状引发的脑网络异常情况,最后基于构建的脑网络参数结合模式识别算法对于预测ADHD疾病的严重程度进行了初步探索。本研究所做的具体工作如下：首先,我们基于96个受试儿童(其中33个ADHD儿童,63个正常儿童)的静息状态fMRI数据,构建每个儿童的复杂脑网络,并求得脑网络参数：三个节点网络参数和六个全局网络参数。研究结果表明儿童脑网络也呈现和成人相似的“小世界”属性,并且有一些与成人类似的对信息传递起重要作用的“hub”脑区。其次,为分析人类注意缺陷严重程度(ADHD的症状之一)个体差异的脑网络机理,我们采用相关分析的方法分析了注意缺陷评分(IAS)与复杂脑网络参数的系数。实验结果表明：双边的尾状核和丘脑、左侧的中扣带和舌回以及右侧的海马的节点网络参数与IAS评分呈现显著的正相关(P0.01),进一步的逐步线性回归分析中,生成的IAS评分预测模型中仅保留了左侧中扣带的中介核心性和左侧尾状核的全局效率。这些研究发现说明这七个脑区的局部网络拓扑组织结构影响了注意行为的个体表现。最后,我们基于复杂脑网络参数采用支持向量回归的方法,对个体注意缺陷水平进行预测。本研究中,我们基于复杂脑网络参数与IAS的相关系数进行特征选择,,分别采用线性核函数和多项式核函数,用支持向量回归的方法构建IAS评分预测模型,通过留一法交叉验证。结果表明,基于与IAS相关性最强的50个特征,采用多项式核函数,可以达到最好的预测效果,预测IAS与真实IAS的相关性系数达到0.44。综上所述,本研究中我们基于静息状态fMRI数据,对ADHD疾病注意缺陷症状的复杂脑网络机理做了一定探索,并对基于复杂脑网络参数开展ADHD注意缺陷症状严重程度预测进行了初步探索。本研究结果不仅有助于推进我们对注意缺陷脑机理的认识,而且对基于磁共振影像的ADHD辅助诊断研究有借鉴意义。
[Abstract]:Attention deficit / hyperactivity disorder (ADHD) is a common mental development disorder in children. In recent years, ADHD has become a serious public health problem in China and the whole world. In recent years, the complex network analysis based on graph theory has greatly promoted the understanding of brain network organization model. In this study, we constructed a complex brain network for every child based on 96 resting state fMRI images, and then analyzed the abnormal brain network caused by ADHD symptoms. Finally, based on the constructed brain network parameters and pattern recognition algorithm, the prediction of the severity of ADHD disease is preliminarily explored. The specific work of this study is as follows: Firstly, based on the resting state fMRI data of 96 subjects (33 ADHD children and 63 normal children), we construct a complex brain network for each child, and obtain the parameters of the brain network: three node network parameters and six global network parameters. The results show that children's brain networks also show similar "small-world" attributes to adults, and there are some similar "hub" brain areas that play an important role in information transmission. Secondly, in order to analyze the brain network mechanism of individual differences in the severity of human attention deficit and one of the symptoms of ADHD, we used the correlation analysis method to analyze the coefficients between the attention deficit score (IASA) and the parameters of complex brain network. The results showed that the nodal network parameters of bilateral caudate nucleus and thalamus, left middle cingulate band and lingual gyrus and right hippocampus were positively correlated with IAS score (P 0.01). Only the mediating core of the left middle cingulate band and the global efficiency of the left caudate nucleus were retained in the IAS score prediction model. These findings suggest that the local network topology of the seven brain regions affects the individual performance of attentional behavior. Finally, based on the parameters of complex brain network, the support vector regression method is used to predict the level of individual attention deficit. In this study, we use linear kernel function and polynomial kernel function to construct the prediction model of IAS score based on the correlation coefficient between the parameters of complex brain network and IAS, respectively, and construct the prediction model of IAS score by using support vector regression method. The results show that based on the 50 features with the strongest correlation with IAS, the best prediction effect can be achieved by using polynomial kernel function, and the correlation coefficient between IAS and real IAS is 0.44. To sum up, based on the rest state fMRI data, we explored the complex brain network mechanism of ADHD disease attention deficit symptoms. The prediction of the severity of attention deficit in ADHD based on complex brain network parameters was explored. The results of this study are not only helpful to advance our understanding of the mechanism of attention deficit brain, but also useful for the study of ADHD aided diagnosis based on magnetic resonance imaging.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP18;R749.94

【共引文献】