新型变分模型的构建及其在脑成像数据中的应用

发布时间：2018-03-10 21:01

本文选题：脑成像数据　切入点：形状重建　出处：《华中科技大学》2016年博士论文　论文类型：学位论文

【摘要】：脑是自然界最复杂的系统之一,支配着人类的思维与行为。研究脑的结构和功能,对描述智力、思维和意识的产生机制,揭示脑的工作原理具有重要意义,进一步能够促进我们对一些高级功能和神经精神类疾病形成机制的理解。随着生命科学、光学、机械和信息科学等多学科的交叉和融合发展,使得在高分辨率水平获取小鼠全脑的数据集成为可能。将获取的脑成像数据转化为生物学知识已成为脑研究的瓶颈问题。作为脑研究的重要部分,脑成像数据的形状重建,在数字化重建神经元、脑成像数据配准、定性定量分析等方面具有重要的意义,为研究神经元形态、定量分析神经元的投射等提供帮助,为探索脑疾病的形成机理奠定基础。然而,由于大的数据量以及高的数据复杂度,使得脑成像数据的形状重建面临巨大的挑战。围绕脑成像数据转化为生物学知识的需求,针对其中形状重建这一具体问题,本文通过构建新型变分模型的方式来进行脑成像数据的形状重建,主要贡献如下：(1)构建了新型的变分模型。在普通变分模型的基础上,本文通过在发射射线采样的信号中给出能量方程的方式构建了新型的变分模型。该模型能够控制边界元的演化方向,解决了普通变分模型在形状重建过程中对初始轮廓的选取比较敏感这一问题;同时,通过控制边界元的演化方向和光滑约束项的引入,解决了噪声信号干扰的问题。(2)基于球坐标系的变分模型,建立了重建神经元胞体形态的方法。该方法将三维图像信号转换到球坐标系中,在球坐标系下构建一个变分模型来重建胞体形态。通过对实际数据的测试,验证了基于球坐标系的变分模型具有重建密集、粗突起干扰等情形下的神经元胞体形态的能力。(3)基于重采样的变分模型,建立了重建鼠脑轮廓的方法。该方法通过重采样的方式获得轮廓附近的局部信号,在重采样的数据中构建一个变分模型来重建鼠脑轮廓。通过对实际数据集的测试,验证了基于重采样的变分模型具有重建数据量大、边界信号不均匀干扰等情形下的鼠脑轮廓形态的能力。本文建立的方法在神经元追踪、数据预处理、鼠脑配准等方面具有实际的应用。重建神经元胞体形态的方法已经应用于密集神经群落和稀疏神经元的追踪方面,通过重建神经元胞体形态,获得与胞体直接相连的树突与轴突的相关信息,为神经纤维的追踪和神经网络的分配提供先验信息;重建鼠脑轮廓的方法已经应用于数据预处理、鼠脑配准等方面,通过获得的脑轮廓,可以去除鼠脑外的干扰信息,使得数据预处理、鼠脑配准更为准确。
[Abstract]:Brain is one of the most complex systems in nature, which dominates human thinking and behavior. It is of great significance to study the structure and function of brain to describe the mechanism of intelligence, thinking and consciousness, and to reveal the working principle of brain. It can further promote our understanding of the formation mechanism of some advanced functional and neuropsychiatric diseases. With the development of life science, optics, machinery and information science, the interdisciplinary and integration of many subjects, such as life science, optics, machinery and information science, It is possible to obtain the whole brain data of mice at high resolution level. Converting the obtained brain imaging data into biological knowledge has become the bottleneck of brain research. As an important part of brain research, the shape reconstruction of brain imaging data, It is of great significance in the digital reconstruction of neurons, the registration of brain imaging data, the qualitative and quantitative analysis and so on, which can be helpful for the study of neuron morphology and the quantitative analysis of neuronal projection. However, because of the large amount of data and high data complexity, the shape reconstruction of brain imaging data faces a great challenge. Aiming at the specific problem of shape reconstruction, this paper constructs a new variational model to reconstruct the shape of brain imaging data. The main contributions are as follows: 1) A new variational model is constructed. In this paper, a new variational model is constructed by giving the energy equation in the radially sampled signal, which can control the evolution direction of the boundary element. The problem that the general variational model is sensitive to the selection of the initial contour in the shape reconstruction process is solved. At the same time, by controlling the evolution direction of the boundary element and the introduction of the smooth constraint term, The problem of noise signal interference is solved. Based on the variational model of spherical coordinate system, a method of reconstructing neuronal cell body morphology is established, which converts 3D image signal to spherical coordinate system. A variational model is constructed in spherical coordinate system to reconstruct the shape of the cell body. Through the test of the actual data, it is proved that the variational model based on the spherical coordinate system has dense reconstruction. Based on the variational model of resampling, a method for reconstructing the contours of the brain is established. The local signals near the contours are obtained by resampling. A variational model is constructed in resampling data to reconstruct the contours of the brain. By testing the actual data set, it is proved that the variational model based on resampling has a large amount of reconstructed data. The ability of brain contours in the presence of inhomogeneous boundary signal interference. The proposed method is used in neuronal tracking, data preprocessing, and so on. The method of reconstructing neuronal somatic morphology has been applied to the tracing of dense nerve communities and sparse neurons by reconstructing neuronal somatic morphology. To obtain the related information of dendrites and axons directly connected to the cell body, to provide a priori information for the tracking of nerve fibers and the distribution of neural networks, the method of reconstructing the contours of the brain has been applied in data preprocessing, registration of the brain, and so on. Through the obtained brain contour, the interference information outside the brain can be removed, which makes the data preprocessing and the brain registration more accurate.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：R338;TN911.7

【相似文献】