【作者】 张波；

【导师】 明东；

【作者基本信息】 天津大学 ， 生物医学工程， 2023， 博士

【摘要】 阈下抑郁症(Subthreshold depression,SD)被视为抑郁症(Major depression,MD)的前驱阶段,尽管未达到MD诊断标准,但仍对个人生活质量和社会医疗资源产生严重影响。目前,SD的临床诊断主要凭借抑郁测评量表和精神科医生的主观评估,且SD与MD具有较高的症状相似性,导致高误诊率问题,加重个人身心和社会医疗负担。因此,临床上亟需客观、精准的生物标记来辅助SD临床诊断。功能磁共振成像技术凭借高空间分辨率、无创性、以及全脑成像等优势为SD生物标记物的探索提供了重要的技术媒介。因此,本研究利用功能磁共振成像技术,从功能活动、功能连接、功能网络的差异化时间与空间视角探索脑功能影像学标记,并开展基于多尺度脑功能特征的影像组学研究进行SD和MD的分类诊断,对于刻画SD脑功能损伤特征、加深对病理机制的理解、推动临床精准诊断的发展具有重要意义。本研究主要工作包括以下四个部分:1.联合差异化空间视角,以体素水平的自发功能活动与网络水平的拓扑特性探索SD脑功能影像学标记。一方面,在体素水平上以低频振荡振幅和局部一致性两种指标评估全脑自发功能活动强度,首次揭示SD患者楔前叶、额中回、额上回、海马区自发功能活动异常。另一方面,设计了网络阈值选择方法以优化功能网络构建,并以图论分析在大尺度脑网络水平探究全局与节点拓扑性质,结果揭示了SD患者功能网络的小世界属性异常,以及楔前叶、额上回、前扣带回、枕叶的节点拓扑异常。该研究实现了SD脑功能异常的高分辨空间定位和大尺度拓扑度量,且体素与网络水平的结果共同表明SD中以默认模式网络(Default mode network,DMN)为主要区域的脑功能损伤,为系统刻画SD全脑功能影像学标记、理解潜在病理机制提供多层次理论基础。2.基于以上发现聚焦于DMN内部,以DMN子区域间功能连接的视角探索SD影像学标记。首先以独立成分分析进行DMN子区域的精细空间划分,之后分别基于各子区域时间序列的相关性、以及滑动时间窗下相关性的波动程度,设计了功能连接强度与波动性分析方法,实现DMN静态与动态特征的全面度量。结果显示SD患者DMN内侧前额叶与后部多区域连接强度的升高,以及海马与楔前叶连接波动性的升高。该研究首次揭示了DMN内部的脑连接影像学标记,为深入理解思维反刍相关的自传体记忆与情绪处理缺陷提供重要的脑损伤基础。3.开展脑功能网络时变共激活模式分析,以更小时间尺度下的网络动态交互视角探索SD脑功能影像学标记。首先提出一种数据驱动的基于预聚类的脑激活判定方法,接着通过共激活模式分析识别出五种跨时间重复出现的脑功能网络,发现SD患者在突显网络的停留时间显著减少,并且更倾向于将大脑由突显网络转换到DMN状态。该研究首次刻画了SD网络时变特性影像学标记,并提出一种支撑DMN功能障碍的潜在动态神经理论,揭示SD表象静态脑损伤的内在动态神经基础,为SD病理机制和脑损伤发展的理解提供动态新视角。4.开展基于多尺度脑功能特征的影像组学研究,探索SD与MD的临床智能诊断分类模型。首先从体素、脑区、脑网络三种空间尺度进行高维脑功能特征的提取,构建了融合多尺度脑功能特征的影像组学分类模型。该模型实现了SD、MD和健康人三分类84.21%的准确率,显著优于现有模型以及基于单一类别特征的分类模型。通过特征分类贡献度分析证明了DMN等脑区在SD与MD分类中的重要地位,印证脑功能影像学标记在个体化精准诊断中的潜在价值,对于推动临床智能诊断的发展以及抑郁障碍病理机制的理解具有重要意义。综上,本研究系统且深入地在多种时间与空间尺度下刻画SD脑功能影像学标记,对于推动SD客观精准诊断的发展具有重要价值。同时,多视角下的脑功能影像学标记为深入理解SD的潜在病理机制,尤其是DMN相关的的情绪与记忆功能异常奠定重要的理论基础。更多还原

【Abstract】 Subthreshold depression(SD)is the precursor of depression,which can not merit the diagnostic criteria for major depression(MD),but it produces significant decrements in personal quality of life and increase in social medical burden.Because the diagnosis of SD is mainly based on depression rating scales and subjective assessment by psychiatrists,and great similarities exist in the clinical symptoms between SD and MD,the clinical diagnosis of SD is facing with a high misdiagnosis rate,which further increase the physical,mental and social medical burden.Therefore,there is an urgent need for objective and accurate biomarkers to improve the clinical diagnosis of SD.With the advantages of high spatial resolution,non-invasiveness,and whole-brain imaging,functional magnetic resonance imaging(f MRI)provides an important tool to explore biomarkers in SD.Therefore,this study would explore the brain functional imaging biomarkers with functional activity,functional connectivity,functional network methods in multiple temporal and spatial scales,and perform multi-scale functional features based radiomics study to classify SD and MD.It is of great significance to facilitate the comprehensive understanding of the brain dysfunction and neuropathology underlying SD,and promote the development of clinical diagnosis.The main work of this study includes the following four parts:First,combining the differentiated spatial perspectives,voxel-level functional activities and network-level topological properties were analyzed to explore the functional imaging biomarkers of SD.On the one hand,amplitude of low frequency fluctuation and regional homogeneity were applied to evaluate the intensity of spontaneous functional activity at voxel-level,and altered spontaneous functional activity in the precuneus,middle frontal gyrus(MFG),superior frontal gyrus(SFG),and hippocampus were found in SD.On the other hand,a novel network thresholding method was designed for functional network optimization,and the global and nodal topological properties of functional network were explored at a larger spatial scale based on graph theory.The results showed altered global small-worldness,and abnormal nodal topology in the precuneus,SFG,anterior cingulate cortex,and occipital gyrus.These results revealed altered functional activities in a high-resolution perspective and abnormal network topology of large-scale networks in SD.The results of both voxel and network-level indicated functional impairments mainly in the default mode network(DMN)in SD.These findings help to establish the functional imaging biomarkers in the whole brain,and provide theoretical basis for the understanding of pathological mechanism of SD.Second,based on the above findings,we further focused on the DMN system and explored the imaging biomarkers of functional connectivity within DMN regions.Independent component analysis was used to accurately define the DMN regions.To assess the static and dynamic characteristics of functional connectivity within the DMN,functional connectivity strength and variability were defined based on correlation of time series and the degree of variability among slide-windows.Results showed that functional connectivity strength between the medial prefrontal cortex(m PFC)and posterior DMN was significantly increased,and the variability between the hippocampus and precuneus was significantly increased in SD.This study reveals the imaging biomarkers of functional connectivity within DMN for the first time,providing an important neural basis for rumination-related memory and emotion processing defects in SD.Third,time-varying co-activation pattern(CAP)analysis was performed to capture the dynamic characteristics of functional network in a smaller time scale and to explore the functional imaging biomarkers of SD.A data-driven pre-clustering based brain activation method was proposed,then CAP analysis was performed and five functional networks that repeated across time points were identified were identified.Results showed that the dwell time of the salience network(SN)was decreased,and the network transition frequency from the SN to DMN was significantly increased in SD.This study characterized the time-varying imaging biomarkers of SD for the first time,and proposed a potential dynamic neural theory supporting the DMN dysfunction.It revealed the intrinsic dynamic neural basis underlying static abnormalities of functional networks in SD,and provided a dynamic perspective for the understanding of pathological mechanism and the development of brain functional abnormalities of SD.Fourth,a radiomics analysis of SD and MD based on multi-scale brain functional features was designed,to explore the intelligent classification model for clinical diagnosis.High-dimensional brain functional features were extracted from voxel,brain region,and network spatial scales,and a radiomics classification model based on fusion features was constructed.The model achieved an overall classification accuracy of 84.21%among SD,MD and HC,which showed significant advantages in classification performance compared with the models based on single-scale functional features and previous studies.In addition,based on the analysis of discriminative power of features,this study further showed the important role of features mainly in the DMN in the classification model,which also proved the potential value of the above imaging biomarkers in individualized diagnosis.These findings was significant to promote the development of clinical intelligent diagnosis and the understanding of the pathological mechanism of depression.In summary,this study systematically characterized brain functional imaging biomarkers of SD at multiple temporal and spatial scales,which was of great value in promoting the development of accurate clinical diagnosis of SD.At the same time,brain functional imaging biomarkers lay an important theoretical basis for a deeper understanding of the pathological mechanisms of SD,especially DMN related emotional and memory dysfunction.更多还原