显微光学切片断层图像预处理方法研究

发布时间：2018-03-28 08:16

本文选题：图像预处理　切入点：伪影去除　出处：《华中科技大学》2013年硕士论文

【摘要】：如何准确认识脑是现代科学研究中一个巨大的挑战，其中神经结构是脑功能与疾病发病机制研究的重要基础。随着显微成像技术的发展，特别是大范围、高分辨光学显微成像技术的出现，不断获取的海量复杂神经解剖结构的图像数据集为脑科学研究提供了有力的工具。面对全脑高分辨成像带来的海量复杂数据，全自动的图像处理方法成为了知识挖掘的必要方法和工具。然而，由于样本制备和成像过程产生的多种伪影会使图像质量下降，因此对海量复杂神经解剖图像处理方法提出了新的需求。基于显微光学切片断层成像系统所获取的数据量达5TB的尼氏染色和8TB的高尔基染色小鼠全脑结构数据集，针对原始切片图像存在的多种伪影，本文提出了一套用于组织染色显微光学图像的全自动伪影去除方法，实现了对包含细胞、血管等结构复杂图集的校正。本文对伪影的去除包括以下几个方面：（1）使用移动中值滤波的方法对图像的均值投影曲线进行平滑，，并使用计算得到的校正系数去除图像中的条纹噪声；（2）使用脑轮廓做掩膜提高非均匀背景提取的准确性，将原图与非均匀背景相减校正染色不均匀导致的图像亮度不均匀；（3）使用形态学滤波的方法对图像中存在的不规则亮斑进行去除；（4）使用分割后脑内包埋剂的均值为参考，将各个断层统一到相同亮度。尼氏染色和高尔基染色小鼠全脑数据集经过本文方法处理后，全脑各区域亮度均匀，图像质量较处理前有明显的提升，校正后的图像能够清晰的展示细胞构筑、血管拓扑和神经细胞形态。亮度均匀的高质量全脑图集可以结合胞体分割、血管追踪和细胞形态检测等方法，用于胞体、血管以及神经细胞类型的定量计算和分析，为神经科学家进一步揭示脑工作机理提供可靠的基础数据集。
[Abstract]:How to accurately understand the brain is a great challenge in modern scientific research, in which the neural structure is an important basis for the study of brain function and the pathogenesis of disease. With the development of microscopic imaging technology, especially in a wide range, With the emergence of high-resolution optical microscopic imaging technology, a large number of image data sets of complex neuroanatomical structures have been continuously acquired, which provides a powerful tool for the research of brain science. In the face of the massive complex data brought by high-resolution imaging of the whole brain, Fully automatic image processing has become a necessary method and tool for knowledge mining. However, because of the variety of artifacts produced in the process of sample preparation and imaging, the image quality is degraded. Therefore, a new demand is put forward for massive complex nerve anatomical image processing. Based on the whole brain structure data set of mice with 5TB and 8TB Golgi staining obtained by the microoptical slice tomography system, a variety of artifacts existed in the original slice images were studied. In this paper, a set of automatic artifact removal methods for tissue staining microscopic optical images is proposed. In this paper, the removal of artifacts includes the following aspects: 1) smoothing the mean projection curve of the image by moving median filter. Using the calculated correction coefficient to remove the fringe noise in the image, the brain contour is used as the mask to improve the accuracy of the non-uniform background extraction. Subtractive correction of the original image with non-uniform background to correct the uneven brightness of the image caused by non-uniform coloring) removal of irregular bright spots in the image by morphological filtering method) use the mean value of the embedded agent in the brain after segmentation as a reference. Unify the faults to the same brightness. After the whole brain data sets of Nieldahl staining and Golgi staining mice were processed by this method, the brightness of all regions of the whole brain was uniform, the image quality was obviously improved compared with that before processing, and the corrected images could clearly show the cellular architecture. High quality global brain atlas with uniform brightness can be combined with cell body segmentation, vascular tracing and cell morphology detection for quantitative calculation and analysis of cell body, blood vessel and nerve cell types. It provides a reliable basic data set for neuroscientists to further reveal the mechanism of brain work.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R310

【共引文献】