【作者】 刘颖；

【导师】 张冰； 徐运；

【作者基本信息】 南京大学 ， 临床医学， 2021， 博士

【摘要】 第一章基于正常对照脑功能连接的模块划分背景:非任务状态下人脑左右运动皮层的血氧水平依赖磁共振成像(blood oxygenation level dependent,BOLD)信号在0.01-0.08Hz频段中显著相关。这反映了在静息态下,大脑的很多区域之间在仍然在进行着信息交流,这种BOLD时间信号的相关性就是功能连接(functionl connectivity,FC)。近年来,研究者将功能连接的研究方法应用于大脑功能研究,并证明了这种静息态下不同脑区之间BOLD时间信号的相关性广泛地存在于不同的脑功能系统内。目的:本部分的实验旨在以对照组的平均功能连接矩阵为依据,基于每个被试的功能连接网络,将大脑划分出功能模块,以便后续的进一步研究。方法:在2016年7月至2017年6月期间,共招募了110名被试,51人因排除标准被排除,最终样本量为59名被试:包括23名遗忘性轻度认知障碍(amnesia mild cognitive impairment,a MCI)患者,18名非遗忘性轻度认知障碍(non-amnesia mild cognitive impairment,na MCI)患者和18名正常对照(normal control,NC)。所有研究对象均来自于南京大学医学院附属鼓楼医院记忆门诊。南京大学医学院附属鼓楼医院伦理委员会批准了研究方案(临床试验政府编号:2016-063-01)。所有患者均签署了书面知情同意书。受试者使用标准化方案进行检查,包括:神经心理筛查、磁共振(magnetic resonance imaging,MRI)和一般医学检查。a MCI的诊断标准如下:主观或客观的记忆损害;记忆减退的客观表现(认知测试):听觉-词语学习测验延迟回忆得分低于年龄和教育匹配对照的1.5个标准差;一般认知功能正常:简易精神状态检查表(mini-mental state examination,MMSE)≥24分;临床痴呆症评定量表(clinical dementia rating,CDR)0.5分;日常生活活动损伤很小或没有损伤;未达到痴呆。对na MCI除了排除了客观记忆损伤的诊断,也使用了类似的标准。排除标准为:脑血管病、严重抑郁即17项汉密尔顿抑郁量表(Hamilton depression scale,HAMD)≥24分、既往脑损伤或者酗酒病史、白质高信号(2级或以上)。对照组受试者没有认知障碍,也没有神经或精神障碍。所有参与者的其他排除标准包括MRI的禁忌症,如心脏起搏器和幽闭恐惧症。所有参与者都是右撇子。磁共振图像采集完成之后,利用Deparsf软件将图像进行校正、去干扰。为了构建个体的FC网络,根据图论理论,先用AAL90模板将大脑分成90个不同的感兴趣区(节点),用节点之间的联系强度来确定感兴趣区域之间的绝对FC强度,从而得到个体的初始FC矩阵。对初始矩阵进行Fisher-Z变换以改善正态性。之后根据对照组的平均FC矩阵,用改进的贪婪优化算法将复杂网络划分成模块。结果:受试者之间在年龄(p=0.147)、性别(p=0.862)、教育程度(p=0.092)和数字广度测试方面没有统计学差异。MMSE、Mo CA、VR(Visual Reproduction,视觉复制)、AVLT(Auditory-verbal learning test,听觉词语学习测试)、AVLT-DR(Auditory-verbal learning test delayed recall,听觉词语学习测试-延迟回忆)、BNT(Boston naming test,Boston命名测试)在三组间均有显著差异(p<0.05)。a MCI组的MMSE、VRDR、VRR、AVLT、AVLT-DR和BNT评分均低于na MCI组和NC组。NC组的Mo CA评分高于两个MCI组(p<0.01)。根据NC组的FC矩阵,健康的大脑网络由四个连接的模块组成。根据它们的解剖位置和前人的研究结果,我们将它们命名为颞枕模块(模块I)、默认模式模块(模块II)、注意模块(模块III)和感觉运动模块(模块IV)。这四个模块分别包含26、28、16和20个解剖区域。结论:基于多模态神经成像技术,我们在健康脑功能网络中识别出四个联系较外界更紧密的模块:颞枕模块(模块I)、默认模式模块(模块II)、注意模块(模块III)和感觉运动模块(模IV)。第二章功能连接、结构连接以及两者相关性对遗忘性轻度认知障碍的预测作用背景:轻度认知障碍(mild cognitive impairment,MCI),被视为阿尔兹海默病(AD)临床阶段的先发阶段。其中,遗忘性轻度认知障碍(amnestic mild cognitive impairment,a MCI)是MCI的重要亚型。a MCI的患者更有可能进展成临床阿尔兹海默病(AD)。目的:本研究的目的是研究两种MCI亚型中的结构和功能连接,并比较结构连接(structural connectivity,SC)-功能连接(functional connectivity,FC)耦合的差异。方法:59名被试包括a MCI、na MCI(non-amnestic mild cognitive impairment,非遗忘性轻度认知功能障碍)患者和健康对照(NC)接受了静息态状态功能磁共振成像(rs-f MRI)和扩散张量成像(DTI)。根据对照组的功能连接网络,将大脑分成多个功能模块后,比较模块中的平均FC和SC以及平均SC-FC相关值;最后利用逻辑回归的方法来查找a MCI的预测指标。并用ROC曲线对预测的效果进行检测。结果:1)三组之间模块水平的平均FC无显著差异。2)在模块I、模块II和模块III中,NC组的FN低于a MCI组na MCI组;而在模块IV中,NC组的FN低于a MCI组;在模块I、III、IV中,NC组的FA高于a MCI组na MCI组;在模块II中,而FA呈现出a MCI<na MCI<NC的趋势,且三组之间两两存在显著差异。3)模块I中NC组的FN-FC相关值高于a MCI组和na MCI组;而NC组的FA-FC相关值低于a MCI组和na MCI组;在模块II、IV中,NC组的FA-FC相关值低于a MCI组和na MCI组。4)模块I中存在FN-FC以及FA-FC趋势线的组间差异。5)Logistic回归分析显示四个模块中FA值均对a MCI具有一定的预测意义(每1%增量,模块II中OR=0.539,p=0.003、模块III中OR=0.805,p=0.031、模块IV中OR=0.551,p=0.004)。模块IV中FN值对a MCI具有一定的预测意义(OR=1.163,p=0.012)。此外,模块II中,FA-FC相关值对a MCI具有一定的预测意义(OR=2.095,p=0.028)、FN-FC相关值对a MCI也具有一定的预测意义(OR=1.205,p=0.026)。ROC曲线进一步验证了不同指标对于a MCI预测的效果。模块I中,FN值的ROC曲线的曲线下面积为0.702;FA的曲线下面积为0.700。模块II中,FA值的ROC曲线曲线下面积为0.818;模块III中,FA值的曲线下面积为0.732;模块IV中,FA的曲线下面积为0.829。此外,在模块I中,FN-FC相关值的ROC曲线曲线下面积为0.668;FA-FC相关值的曲线下面积为0.673;在模块II中,FA-FC相关值的ROC曲线曲线下面积为0.713;在模块IV中,FA-FC相关值的ROC曲线曲线下面积为0.688。结论:模块水平的结构连接在AD或a MCI进展中可能存在代偿;模块水平的结构连接参数以及结构-功能相关性参数可能对a MCI的预测有意义,还有待进一步验证。更多还原

【Abstract】 Chapter 1 Module division based on normal brain functional connectivityBackground: Functional connectivity(FC)refers to the level of resting state blood oxygen in the left and right cortex of the human brain.In recent years,researchers have applied this method of FC to multiple brain functional systems,and proved that the correlation of BOLD signals between different brain regions in this resting state widely existed in different functional systems.Objective:The purpose of this part of the experiment was to divide the brain into functional modules based on the FC network of the normal control(NC)group for further research.Methods: During July 2016 to June 2017,a total of 110 subjects were recruited and 51 were excluded due to the exclusion criteria.The final sample size was 59,including 23 patients with amnesia mild cognitive impairment(a MCI),18 patients with non-amnesia mild cognitive impairment(na MCI)and 18 normal controls(NC).All subjects were from the memory clinic of Affiliated Drum Tower Hospital,Nanjing University.The ethics committee of Drum Tower Hospital approved all the research procedure(clinical trial government code: 2016-063-01).All patients signed a written informed consent form.The subjects underwent a standardized program for examination,including neuropsychological screening,MRI,and general medical examination.The diagnostic criteria of a MCI were as follows: subjective or objective memory impairment;objective memory performance documented by an Auditory-Verbal Learning Test delayed recall(AVLT-DR)score ≤1.5 standard deviations of age-adjusted and education-adjusted norms;normal general cognitive function: mini mental state examination(MMSE)≥ 24;Clinical dementia rating scale(CDR)equaling 0.5;little or no daily life activity damage;not reaching the diagnosis dementia.Similar criteria were used for classifying na MCI,but with objective memory impairment as an exclusion criterion.The exclusion criteria were as follows: cerebrovascular disease;severe depression(17 items of Hamilton Depression scale(HAMD)≥ 24),previous brain injury or alcoholism history;white matter high signal(WMH,no less than grade 2).The subjects in the normal control group had no cognitive impairment and neurological or mental disorders.Other exclusion criteria for all participants included contraindications for MRI,such as pacemakers and claustrophobia.All the participants were right-handed.After the MRI data were preprocessed,in order to construct a weighted FC network,weighted edges were used to determine the absolute FC intensity between connected regions of interest.Then the correlation matrix is transformed by Fisher-Z transformation to improve the normality of the correlation coefficient.Then the complex network was divided into different modules through improved greedy optimization algorithm.Results: No statistically significant differences among subjects was found in age(p = 0.147),sex(p = 0.862),education(p=0.092),or the Digital Span test.Significant differences in the MMSE,Mo CA,VR subtest,AVLT,AVLT-DR,and BNT were found among the three groups(p<0.05).The a MCI group had lower MMSE,VRDR,VRR,AVLT,AVLT-DR,and BNT scores than both the na MCI and the NC group(p<0.05).The NC group had higher Mo CA scores than both MCI groups(p<0.01).According to the FC matrices of the NC group,the healthy functional brain network comprised four connected modules.According to their anatomical locations and findings from previous studies,we labelled them the temporaloccipital module(Module I),the default mode module(Module II),the attention module(Module III),and the sensorimotor module(Module IV).These four modules contained 26,28,16,and 20 anatomical regions,respectively.Conclusions: Based on multimodal neuroimaging technology,we identified four interconnected modules in the healthy functional brain network,namely,the temporooccipital lobe module(module I),the default mode module(module II),the attention module(module III)and the sensorimo tor module(module Ⅳ).Chapter 2 The predictive effects of functional connection,structural connection and their correlation on amnesia mild cognitive impairmentBackground: Mild cognitive impairment(MCI),regarded as the prodromal stage before the clinical phase of Alzheimer’s disease(AD),has been considered for early differential diagnosis.Patients with amnestic mild cognitive impairment(a MCI),an important subtype of MCI,are more likely to progress to clinical Alzheimer disease(AD).Objective:The aim of this study was to examine the structural and functional connectome in both subtypes of MCI and found differences in structural connectivity(SC)-functional connectivity(FC)coupling.Methods: Fifty-nine individuals with aMCI,naMCI and healthy controls(NC)underwent resting-state functional magnetic resonance imaging(rs-f MRI)and diffusion tensor imaging(DTI).Brains were parcellated into functional modules according to the FC networks of NC,and individual FC and SC within modules as well as average FC,average SC and average SC-FC correlation values in each module among groups were analyzed,while logistic regression was used to find predictive factors.The effect of prediction was tested by ROC curve.Results: 1)No significant differences were found in the average FC of the four modules among the three groups.2)In module I,module II and module III,the FN value of NC group was lower than that of a MCI and na MCI group,while in module IV,the FN of NC group was lower than that of a MCI group;in module I,III and IV,the FA of NC group was higher than that of a MCI group na MCI group;in module II,FA showed a trend of a MCI < na MCI < NC,significant differences were found between every two groups among the three groups.3)In module I,the FN-FC correlation value of NC group was higher than that of a MCI group and na MCI group,while the FA-FC correlation value of NC group was lower than that of a MCI group and na MCI group;in module II and IV,the FA-FC correlation value of NC group was lower than that of a MCI group and na MCI group.4)There are inter-group differences of FN-FC and FA-FC correlation curves in module I.5)Logistic regression analysis showed that FA values in the four modules all had predictive significance for a MCI(per 1%increment,OR=0.539,p=0.003 in module II;OR=0.805,p=0.031 in module III;OR=0.551,p=0.004 in module IV).The FN value in the module IV had a certain predictive significance for a MCI(OR=1.163,p=0.012).In addition,in the module II,the FA-FC correlation value had a certain predictive significance for a MCI(OR=2.095,p=0.028),and the FN-FC correlation value also had a certain predictive significance for a MCI(OR=1.205,p=0.026).The ROC curve further validated the effect of different indicators on a MCI prediction.In module I,the area under the curve(AUC)of the ROC curve of FN value was 0.702.The AUC of FA was 0.700.In the module II,the AUC of FA value was 0.818;in the module III,the AUC of FA value was 0.732;in the module IV,the AUC of FA was 0.829.In addition,in module I,the AUC of FN-FC correlation value was 0.668;in the module II,the AUC of FA-FC correlation value was 0.713;in the module IV,the AUC of FA-FC correlation value was 0.688.Conclusions: Module-level structural connectivity may be compensated in the progress of AD or a MCI;module-level structural connection parameters and structure-function correlation parameters may be meaningful for the prediction of a MCI,which need to be further verified.更多还原