基于线粒体结构的神经组织显微图像配准技术研究

CASIA OpenIR > 毕业生 > 硕士学位论文

	基于线粒体结构的神经组织显微图像配准技术研究
	曹会文
	2017-05
学位类型	工程硕士
中文摘要	逆向构建纳米尺度脑神经回路物理结构，其目的是为探求神经回路结构与大脑功能之间的关系提供依据。目前国际上主要的研究机构采用序列切片电镜成像的方式来获取纳米尺度脑神经回路物理结构，该方法涵盖样品制备、切片、成像、图像配准、三维重建等环节。本文研究的神经组织序列切片显微图像配准方法，是纳米尺度脑图谱构建过程中的一个关键环节，目标是恢复神经组织样本的三维连续性和几何特性，为后续重建分析提供良好的三维图像数据集。但神经组织序列图像配准面临各种难题：在样品制备、染色、切片、成像过程中不可避免的引入图像畸变，如旋转平移、污染、组织形变等等；就切片本身来说，虽然相邻切片间的图像内容具有一定的相似性，但其相似程度取决于样品组织的局部结构和切片厚度，切片越厚其相似性越低；此外，电镜图像分辨率可达到几纳米，而切片厚度为几十纳米，因此图像在三维尺度上表现为各向异性。上述种种因素为神经组织序列显微图像的配准带来了很大挑战。目前的图像配准算法大多是基于传统图像特征进行配准，本文提出一种基于生物学识别特征的图像配准方法，矫正序列切片在收集、成像过程中引入不可控的形变，将序列切片图像约束在可更精确配准的范围，减少配准过程中的人工交互。我们选取线粒体中心点作为图像的配准点，主要基于以下原因：线粒体的广泛分布保证我们能够在整幅图像各个位置获取配准点信息；线粒体在切片上呈现的相似形状便于我们进行检测并提取特征；线粒体长度一般在 1 微米至 2 微米之间， 70 纳米的切片厚度能够保证我们可以在连续的十几张切片图像中获取相同的特征；此外，在切片厚度为几十纳米时，上下层切片上线粒体形态和大小可能会有一定的变化，但线粒体中心点的位置基本保持不变，这将减小切片厚度对图像配准精度的影响。因此，选取线粒体中心点作为图像的配准点解决了神经组织序列显微图像配准点的获取难题，这也正是本文的创新之处。本文提出的基于线粒体结构的神经组织显微图像配准方法主要分为基于线粒体的对准点提取、对准点匹配和基于控制点的神经组织显微图像变形等步骤。本文选取果蝇蘑菇体序列切片电镜图像作为实验数据，应用基于线粒体结构的方法进行配准，实验结果表明该方法能够很好地将神经组织显微图像进行配准，并且保持了生物组织形态。
英文摘要	Reverse engineering the physical structure of nanoscale brain neuron circuit aims to explore the relationship between the neuron circuit and brain function. At present the main international institutions obtain the physical structure of nanoscale brain neuron circuit with the method of serial section Electron Microscope imaging. This method includes sample preparation, sample cutting, imaging, image registration and 3D reconstruction. Microscopic image registration is a key component of the neural structure reconstruction with serial sections of neural tissue. The goal of microscopic neural image registration is to recover the 3D continuity and geometrical properties of specimen. During image registration, various distortions need to be corrected, including image rotation, translation, tissue deformation et.al, which come from the procedure of sample cutting, staining and imaging. Furthermore, there is only certain similarity between adjacent sections, and the degree of similarity depends on the local structure of the tissue and the thickness of the sections. Besides, the X-Y resolution of microscopic image can achieves 4 nanometer, but the section thickness is dozens of nanometers, which lead to the anisotropic 3D reconstruction. These factors make the microscopic neural image registration a challenging problem. Most of the current image registration algorithms align images based on the conventional image features. In this paper, an image registration method based on biometrics features is proposed. The method could correct serial sections with uncontrolled deformation drived from the collection and imaging process, restrict the serial sections image within the scope in which the serial sections image could be more precisely registered. And it reduces the manual interaction during the registration process. Here are the reasons of choosing the center of mitochondrion as landmarks for image registration. Mitochondrion are widely distributed in the neural tissue image, which guarantees that we can achieve broad distribution landmarks; the ellipsoidal shape of mitochondria ensures that the same mitochondria has similar shape on serial sections, which is beneficial to detect and extract the landmarks; the mitochondria are typically between one and two micrometers in length, and the 70 nanometers section thickness assures that a mitochondria may appear in more than ten sections, thus we can acquire continuous feature in serial sections. Besides, the centers of mitochondrion remain unchanged, reducing the effect of section thickness for registration accuracy. Therefore, choosing the center of mitochondrion as landmarks for image registration to solve the difficulty of corresponding landmarks extraction is the innovation of this article. The proposed image registration method contains three parts: landmarks extraction from adjacent sections, corresponding landmarks matching and image deformation based on the correspondences. We demonstrate the performance of our method with SEM images of drosophila brain. The result shows that the registration was successful and the morphology of biological tissue was maintained.
关键词	神经组织显微图像线粒体特征提取点集匹配图像变形
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/14632
专题	毕业生_硕士学位论文
作者单位	中国科学院自动化研究所
第一作者单位	中国科学院自动化研究所
推荐引用方式 GB/T 7714	曹会文. 基于线粒体结构的神经组织显微图像配准技术研究[D]. 北京. 中国科学院研究生院,2017.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
学位论文_曹会文.pdf（6340KB）	学位论文		限制开放	CC BY-NC-SA