【作者】 张晨；

【导师】 袁菁；

【作者基本信息】 华中科技大学 ， 光学工程， 2020， 博士

【摘要】 大脑作为人类身体中最为复杂和重要的器官,决定了思维和行为。数量众多的神经元是大脑运行的基本结构单元,不同类型的神经元分布不同,形态各异,在大脑中承担的功能也各不相同。在全脑范围内对特定类型神经元进行分布研究不仅是理解大脑运行机制的基础,还对大脑的生理学活动、病理学改变和药物治疗的研究具有重要意义。由于缺乏合适的研究工具,在全脑范围内特定类型神经元精确的细胞个数和密度信息这一基本而重要的问题仍然不得而知。因此,如何更加高效地获取完整的全脑数据集并在此基础上进行全脑特定类型神经元的定量分布研究是目前神经生物学研究中亟待解决的重要技术瓶颈之一。针对这一需求,本文提出了一种利用高分辨率全脑光学成像自动获取并进而定量分析小鼠特定类型神经元全脑分布的技术方案。本文从全脑光学成像数据采集过程的优化、全脑细胞分布三维数据集定量分析平台的构建和特定神经元全脑分布的示范性应用三个方面,展开了系列工作。本文对结构光照明荧光显微光学切片断层成像技术进行了针对性流程优化,系统分析了数据采集过程的流程与耗时,改进了全脑光学成像过程中的图像压缩策略。利用生物样本与包埋介质的特性差异,开发了基于边界自动识别的成像区间自动修改功能,有效减少了冗余数据并缩短了TB量级数据集的获取时间。优化数据管理及实时上传策略,避免了数据传输所需的额外上传时间。以上工作使得以亚微米像素分辨率采集小鼠全脑三维数据集时间缩短将近一半,减少了人工干预,实现了快速稳定获取小鼠全脑数据集。基于以上改良的高分辨率全脑光学成像系统与自动化细胞识别算法相结合,本文发展了一套用于全脑神经元分布精确定量分析的研究流程。定量对比了传统二维细胞识别方法和三维细胞识别方法的差异,证明了由全脑光学成像所支持的三维细胞识别方法在统计小鼠全脑神经元数目方面准确性更高。并分析了成像分辨率对于胞体识别准确率的影响。在典型数据中评估对比了Neuro GPS、FARSIGHT和TCL三种胞体自动识别算法的适用性,选择Neuro GPS算法进行本文全脑细胞分布自动分析。进一步地,建立了完整的小鼠全脑特定类型神经元分布的研究流程,包括小鼠全脑数据集的冠状面图像预处理、鼠脑在三维空间中的方位校正、脑区划分与空间定位、神经元胞体的三维重建与形态学参数的获取等关键环节。利用上述成像系统和分析方案,以分泌促生长激素抑制素释放激素(Somatostatin,SOM)的神经元为例,开展了小鼠全脑特定类型神经元定量分布的示范性研究。获取了SOM-IRES-Cre;Ai3转基因小鼠全脑数据集。对全脑范围内的所有被标记的SOM神经元进行了三维定位和识别,首次获取了小鼠全脑SOM神经元的精确数量。对SOM神经元主要分布的18个脑区进行了人工划分,进而获得了不同脑区中SOM神经元的定量分布情况,并进行了统计学对比分析。进一步地,获取了不同脑区中SOM神经元胞体的形态学信息,并进行了统计学差异分析,为神经元细胞分类问题提供了重要的形态学参考。该研究有望提示SOM神经元在不同脑区参与的脑功能研究及相关神经系统疾病的机制研究。本论文通过改进高分辨率全脑光学成像并与三维立体细胞识别方法结合,建立了一套小鼠全脑特定类型神经元分布的研究方法,并应用此方法获取并定量分析了SOM神经元在小鼠全脑中的分布情况,展示了全脑光学成像在解析全脑神经精细结构中的巨大潜力,证明了该分析平台有望成为特定细胞类型研究的有力工具。更多还原

【Abstract】 Brain is the most complex and important organ in human’s body and determines mind and behavior.A large number of neurons are the basic structural units of the brain function.Different types of neurons have different distributions,different shapes,and different functions in the brain.The study of the brain-wide distribution of type-specific neurons is not only the basis for understanding the mechanism of brain function,but also important for the study of physiological activities,pathological changes and drug treatment of the brain.The precise cell number and density of type-specific neurons in the entire brain remain unknown because of a lack of suitable research tools.Therefore,how to obtain a complete data set of the whole brain more efficiently and further facilitate the researches on the distributions of type-specific neurons is an important problem to be resolved.Thus,this dissertation proposes a technical solution for automatically acquiring and analyzing the distribution of type-specific neurons in the whole mouse brain.This dissertation carries out systematic works from three aspects: optimizing the whole-brain optical imaging of data acquisition process,developing a three-dimensional quantitative analysis platform for brain-wide cell distribution,and quantifying the brain-wide distribution of SOM neurons as an exemplary application.In this dissertation,the data acquisition process of Structured Illuminationfluorescence Micro Optical Sectioning Tomography(SI-f MOST)was optimized specifically.This dissertation improved the image compression strategy of the imaging system through the time-consuming analysis of the system operation process.Automatic identification of biological sample boundaries was developed to reduce the data acquisition time and avoid the data redundancy of the TB-level data set.Data management and realtime upload strategy were optimized to avoid additional upload time required.The above work had shortened the time to obtain the three-dimensional data set of the whole mouse brain with sub-micron pixel resolution by nearly half,and the goal for fast and stable acquisition of whole mouse brain data sets is achieved.The combination of above mentioned whole-brain imaging system with automated cell recognition algorithm,the dissertation proposed a research process for a precise quantitative analysis platform for brain-wide neuronal distribution.This dissertation quantitatively compared the differences between traditional two-dimensional and novel three-dimensional cell recognition methods.The results demonstrated that the three-dimensional cell recognition method supported by whole brain optical imaging has higher accuracy.And the influence of imaging resolution on the accuracy of soma recognition was analyzed.Further,after quantitatively evaluating three candidate algorithms(Neuro GPS,FARSIGHT and TCL)in typical data.Neuro GPS was selected as the cell counting method adopted in this dissertation.Further,a complete research process for the distribution of type-specific neurons in the mouse whole brain was established,including coronal plane images preprocessing,orientation correction,brain region division,and cell body morphological parameters acquisition.To evaluate this above analysis platform,this dissertation studied the quantitative distribution of Somatostatin neurons in the whole mice brains.Three datasets of SOMIRES-Cre;Ai3 transgenic whole mice brains were obtained.Three-dimensional automatic recognition and stereo recognition of SOM neurons in each whole brain were performed,and the precise numbers of SOM neurons in the whole mice brains were obtained for the first time.The 18 main brain regions of the three datasets were divided,and the quantitative distributions of SOM neurons in different brain regions were obtained and corresponding statistical analysis was performed.Further,the morphology of SOM neuron cell bodies in different brain regions were quantitatively analyzed,and statistical differences were analyzed.In summary,this dissertation proposed a research scheme of counting type-specific neurons in whole mouse brain dataset through combining improved high-resolution wholebrain optical imaging and three-dimensional cell recognition method.The dissertation applied this scheme to obtain the barin-wide distribution of SOM neurons quantitatively.The results demonstrated that this scheme for the brain-wide cell distribution has potential to become a powerful tool for type-specific neurons research.更多还原