【作者】 杨静；

【导师】 龚启勇； 雷都；

【作者基本信息】 四川大学 ， 影像医学与核医学， 2021， 硕士

【摘要】 目的:特发性震颤(essential tremor,ET)是最常见的运动障碍疾病之一。临床主要表现为双侧上肢姿势性或运动性震颤,可能伴有其他神经系统软体征或非运动症状表现。近年来,随着图论分析的快速发展,利用图论探索大脑网络拓扑组织异常已成为多种神经精神疾病的重要研究方向。目前,在特发性震颤中已有研究利用静息态功能磁共振成像数据构建脑网络进行图论分析研究,揭示了静息态下ET患者脑网络的拓扑属性改变。而现阶段对ET患者基于磁共振结构影像的脑灰质形态学网络研究十分有限,限制了我们从结构模态的角度对ET的发生发展机制进行综合全面的理解。因此,本研究侧重于利用高清磁共振结构成像序列,结合图论分析手段对未用药ET患者的脑灰质形态学网络的拓扑结构及其与临床症状之间的关系进行探索。同时为了探究灰质形态学网络特征对于早期ET诊断是否存在临床价值,我们将进一步利用机器学习技术将ET患者和健康对照(healthy control,HC)进行分析归类。材料和方法:研究共纳入了36例未用药ET患者,以及37例HC,所有符合条件纳入的样本均完成了高清磁共振结构序列扫描。依托MATLAB 2013b的程序环境,利用主流的医学图像处理软件SPM12对原始影像数据进行预处理。基于上述预处理的结果,利用王金辉团队提出的个体灰质网络的构建方法提取所有被试大脑的全脑灰质形态学网络连接矩阵,然后利用GRETNA软件(图论分析工具)来分析每个被试的个体灰质形态学脑网络的拓扑学属性参数。最后,使用置换检验统计分析两组被试的拓扑属性;确定组间差异拓扑属性参数后,进一步评估上述差异拓扑属性参数和ET患者临床症状评分之间的相关性。另一方面,将提取的全脑灰质形态学网络连接矩阵作为特征,基于支持向量机构建机器学习模型来对ET患者和健康对照进行分类。结果:病例组和健康对照组的脑灰质形态学网络均显示出小世界属性,但是两组之间存在显著差异,包括:(1)在全局拓扑属性层面上,与对照组相比,未用药ET患者脑灰质形态学网络的全局效率增高,特征路径长度降低;(2)在节点拓扑属性层面上,ET患者存在多个脑区的节点中心性改变,同时涉及运动区域(震颤网络:小脑-丘脑-运动皮层网络)和非运动区域(默认模式网络);(3)ET患者中异常改变的左侧小脑节点效率与震颤评分呈负相关(p=0.02,r=-0.41);右侧尾状核节点度分别与汉密尔顿抑郁量表(p=0.04,r=-0.36)以及汉密尔顿焦虑量表(p<0.01,r=-0.36)呈负相关;(4)基于全脑灰质形态学网络连接矩阵,利用支持向量机将ET患者和HC进行分类,结果显示准确率可以高达80%。结论:基于磁共振结构成像结合图论分析方法,我们发现未用药ET患者全脑灰质形态学网络趋于更加随机化模式,反映在脑网络的信息处理能力以及远程传输能力均下降;同时也存在多个拓扑属性异常改变的脑区伴随包括震颤网络、默认网络在内的网络连接异常,涉及到运动和非运动系统的功能损伤。同时,结合机器学习,验证了有关灰质形态学网络的诊断价值。另一方面,在本研究中,我们回顾了之前的有关ET患者的脑网络连接研究,将结果进行整合讨论分析,进一步探索了关于ET的脑网络损坏机制,夯实了基于灰质形态学网络的图论分析影像指征,为之后对该疾病更深一步的诊断和治疗提供了一定意义上的理论支持。更多还原

【Abstract】 Objective:Essential tremor（ET）is one of the most common movement disorders.The main clinical symptoms include abnormal postural/kinetic tremor of bilateral upper limbs,often accompanied by other neurological soft signs or non-motor characteristics.Recent years,with the rapid development of graph theory in neuroimage,the analysis of brain network topological organization has become an important research direction for many neuropsychiatric diseases.At present,the functional network topological properties have been studied using resting-state MRI（RS-f MRI）in patients with ET,revealing the alterations in the topological properties of functional brain network.However,the researches on gray matter（GM）network of ET patients based on structural MRI are very limited,which limits our comprehensive understanding of the occurrence and development mechanism of ET.Therefore,in current research,we studied abnormal topological structure of the GM network in drug-naive patients with ET using high-resolution MRI and graph theory techniques.For further exploring the translational possibilities of individual GM morphological network features in helping clinical early-stage ET diagnosis,we also analyzed the discriminative abilities of individual network information using machine learning methods.Materials and Methods:A total of 36 ET patients and 37 healthy control（HC）were included in this study.All participants underwent high-definition magnetic resonance structural sequence scanning,and the raw images were preprocessed by SPM 12 toolbox based on the MATLAB（version 2013b）programming environment.The GM morphological network matrix of each subject was extracted using the method proposed by Wangjinhui and then GRETNA was used to calculate the topological measurements of the GM morphological networks.Finally,the network topology measurements of the two groups were statistically analyzed by permutation test.After determining the different topological measurements between groups,the correlation between the above altered topological measurements and clinical symptom scores of ET patients was further evaluated.The discriminative abilities of individual GM morphological network matrixes in separating the patients were evaluated using the support vector machine（SVM）algorithm to build machine learning model.Results:The patients with ET showed significantly different topological properties compared with the healthy subjects including（1）higher global efficiency(E_glob)and lower characteristic path length（L_p）in patients with ET at the global network level;（2）abnormal topological centralities involving motor regions（tremor network:cerebellar-thalamic-cortical network）and non-motor regions（default mode network,DMN）in patients with ET at the nodal level;（3）correlation analyses showed that the abnormal topological measurements in patients with ET were correlated with Fahn–Tolosa–Marìn Tremor Rating Scale（p=0.02,r=-0.41）,Hamilton Depression Scale（p=0.04,r=-0.36）and Hamilton Anxiety Scale（p<0.01,r=-0.44）;（4）using the whole brain morphological network metrics,the classification between ET patients and HC based on SVM can reach an accuracy rate of up to 80%.Conclusion:Based on the structural MRI data combined with graph theory,we found that the GM morphological network tended to be more randomized in patients with ET.It was reflected in the decrease of information processing ability and remote transmission ability of brain network.There were also several brain regions with abnormal topological properties and associated with abnormal network connectivity,including tremor network and default mode network,which involved functional impairment of both motor and non-motor brain areas.Combined with machine learning,the diagnostic value of GM morphological network was verified.Finally,we reviewed the previous studies on the brain network connectivity of ET,discussed and analyzed the results,further explored the mechanism of brain network of ET,consolidated the graph theory analysis of image indications based on structural network,and provided theoretical support for further diagnosis and treatment of ET.更多还原