【作者】 熊光磊；

【导师】 季梁；

【作者基本信息】 清华大学 ， 控制科学与工程， 2006， 硕士

【摘要】 细胞是生物体的基本组成部分,它的结构和功能对生命活动具有重要意义。细胞不仅是大量化学物质的容器,也是膜、泡囊和信号传导通路组成的复杂系统,尽管通过一个多世纪的努力,细胞生物学研究取得了若干重大进展,但是关于细胞内特定高分子(如蛋白质、蛋白质复合体、核酸和信号传导物质等等)的空间组织和运动规律以及组织内细胞空间关系等许多问题的机理目前仍不清楚。了解复杂的生命过程需要将不同层次的信息和知识加以整合,基因组和蛋白质组的研究目前很热门并取得了重要进展,基因组学提供了基因层次上的研究工具,而蛋白质组学探讨了蛋白质的结构、功能和相互作用。这些成果为了解细胞提供新途径和手段的同时,也使人们意识到研究仅针对基因和蛋白质很多情况是不够的,继续在细胞层次上的深入探索是必要的和迫切的。近年来,随着细胞研究的需求日益提高,新的细胞成像技术也迅速发展,趋势主要表现在从细胞整体结构到亚细胞结构,从瞬间观察到连续观察,从二维单通道到三维多通道,从体外到活体,从无标记到有标记。在众多的成像手段中,本论文将以一种新兴的技术—光学相干层析为背景,首先简单介绍了OCT的背景、基本原理和发展方向,然后着重叙述了我们自己的OCT系统的搭建和调试过程,最后提出了一个全新的OCT理论模型—PFMC,它能够解释实验中观察到的OCT信号指数衰减现象。各种各样的成像手段让人们观察细胞变得容易的同时,产生的图像数据也给分析处理带来新的挑战。通过观察可以定性地评估细胞图像,然而,人工地从这些图像获取有用信息则是一项非常繁重和耗时的任务,而且结论往往是不可重复、主观的,自动图像分析则能够弥补这些不足。本论文以显微图像中细胞跟踪、荧光图像中的神经元细胞标记和果蝇细胞分割为对象,提出了基于动态高斯混合模型和基于数学形态学的适合不同形状细胞的两种跟踪算法、基于多尺度的曲线结构检测的神经元自动标记算法、基于Level set和Voronoi图的两种果蝇细胞RNAi图像自动分割算法。这些方法具有成本低、速度快、可重复和客观等优点,我们希望它们的引出对相关细胞生物学研究有所帮助。更多还原

【Abstract】 The cells are the building block of organisms. Their structure and function are of fundamental importance to many aspects of biological processes. A cell is not just a well-mixed container of chemicals, but an extremely intricate system of membranes, vesicles and signaling pathways. Although a number of milestones have been built though endeavor for over a century, much is still left to discover about the spatial organization and mobility of macro-molecules (such as proteins, protein complexes, nucleic acids, signaling substances, etc.) in cells, and the spatial relationships between cells in tissues. Understanding complex biological systems requires information and knowledge from many levels. Genomics and proteomics are recently popular and significant progresses have been made. Genomics provides important tools for studies of individuals on the genetic level, whereas proteomics explores structure, function and interactions of proteins. Although their discoveries provide tools and paths for the further understanding of cells, it is recognized that in many cases focusing only on gene and protein is inadequate; advancing more widely and deeply in the level of cell is necessary and imperative.In recent years, with the increasing demand of cell research, cellular imaging technologies develop rapidly. The trend is in ways: from observing whole cells to sub-cell structure, from observing statically to observing lively, from two-dimensional and single channel to three-dimensional and multiple channels, from in-vitro to in-vivo, as well as from no labeling to labeling. Among many imaging methods, we choose a novel one, Optical Coherence Tomography as our basis. The background, principle and advances of OCT are briefly introduced at first. Then, we describe the procedures of constructing and debugging our OCT system in details. Finally, a novel theoretical model named as PFMC is presented. It explains correctly the fact that the OCT signal decays exponentially penetrating into the sample.Whilst the observation of cells is routine with the help of varieties of imaging techniques, large sets of image data pose new challenges to the processing and analysis of them. Although the qualitative evaluation of cell images can be preformed visually, the manual interpretation of them is a tedious and time-consuming task. Moreover, irreproducible and subjective conclusions are usually drawn. Automatic image analysis is able to overcome these shortcomings. The thesis considers three topics: cell tracking in microscopic images, neurite labeling and Drosophila cell segmentation in fluorescence images. We propose two automated methods for tracking of cells in different shapes. One is based on dynamical Gaussian mixture model and the other is based on mathematical morphology. The neurite labeling is resolved by our novel multi-scale curvilinear structure detector. Two methods are utilized to segment the Drosophila cell images from RNAi experiments. They are based on Level set and Voronoi diagram, respectively. These approaches have the advantages of low cost, fast speed, reproducibility and objectivity. We hope they can serve as candidate tools for cell research.更多还原