基于超像素方法的腹部医学影像多目标分割及脏内器官三维重建研究

发布时间：2018-05-04 00:18

本文选题：腹部医学影像 + DICOM　；参考：《吉林大学》2017年硕士论文

【摘要】：医学影像是医学诊断领域一项创新性发现,医学影像的分割处理和三维重建技术在医学研究和临床医学等领域中扮演着非常重要的角色。通过对医学影像处理和三维建模,医生不需要进行解剖手术就可以多角度、多形式观察和诊断病变区域。这种革命性的新技术改变了传统的医学研究模式,为疾病的准确诊断提供了更好保障。本文从腹部二维医学影像出发,介绍了医学影像分析与处理直到器官三维重建的整体过程。医学影像与通用图像有所不同,需要经过格式转换并准确而清晰的在个人计算机中显示,格式转换是其他医学影像研究的基础内容。一幅典型的医学影像是十分复杂的,可能包含了许许多多的器官和组织,在成像过程中由于成像设备和人为因素等影响常常会产生或多或少的干扰信息,医学影像预处理过程看似无关紧要但却可以起到很好的辅助作用,为其他算法的效果质量和可用性提供了支持。根据各种医学影像的成像原理和成像特点,可能并不是所有的信息都是对医疗诊断有帮助的,有时对感兴趣组织和器官的医学影像分割显得尤为重要,为临床医学的研究和发展有非常积极地影响。另外,根据二维切片序列重建器官和组织的三维模型在医学影像分析领域应用价值非常广泛,能够重建出各种器官的三维立体模型,对组织病变预测和人体内部结构探查具有很好的协助作用。不仅可以提高诊断精确性,同时也为医生和相关学者提供了全方位的、立体的器官模型,方便了研究和深度剖析。本文从研究目的出发,首先介绍了DICOM标准,然后将医学影像转换成位图图像,并采用与设备无关的显示模式对位图进行显示。根据信息的重要程度有选择的读取DICOM文件头中与患者、设备以及图像参数等相关的重要信息。根据医学影像的分辨率不高、对比不明显等特点,本文对图像预处理方法做了针对性的设计。具体包括:灰度变换和图像增强以及常用的包括平移的多种几何变换方式,而图像增强又分为平滑、锐化等操作。对比传统的医学影像分割方法,本文提出了一种从腹部医学影像数据中自动分割多个器官的方法,研究目的是通过构建超像素并用其预测更准确的分割。基于超像素的分割方法适应CT图像中的各种成像条件,并且考虑了多个器官之间的相互关系。该方法的框架如下:(1)根据像素相关性和位置邻近性对超像素进行聚类(2)引入器官空间结构分布图,修正多个器官的分割过程。在我们的框架中,通过使用常规单器官分割方法分割足够精确的预测器官,并且使用超像素聚类方法分割剩余器官。对分割结果进一步研究,在腹腔二维切片感兴趣器官分割的基础上,对Marching Cubes(MC)方法进行研究,算法关键点和难点方面主要包括确定体素与等值面相交关系、以及求出此时等值面的法向量。最后给出了腹腔内特定器官经过重建之后的三维模型。
[Abstract]:Medical image is an innovative discovery in the field of medical diagnosis. Medical image segmentation and 3D reconstruction play a very important role in medical research and clinical medicine. Through medical image processing and three-dimensional modeling, doctors do not need to carry out anatomic hand surgery for multiple angles, multiple forms of observation and diagnosis of diseases. This revolutionary new technology has changed the traditional medical research model and provided a better guarantee for the accurate diagnosis of the disease. This paper introduces the overall process of medical image analysis and processing until the three-dimensional reconstruction of the organs from the two-dimensional medical image of the abdomen. A typical medical image is very complex and may contain a lot of organs and tissues. In the process of imaging, the imaging equipment and human factors often produce more or less interference letters in the imaging process. Medical image preprocessing seems to be insignificant but can play a very good auxiliary role in supporting the quality and availability of other algorithms. According to the imaging principles and imaging features of various medical images, not all information is helpful to medical diagnosis, and sometimes it is of interest to organizations and organs. The medical image segmentation is particularly important, and it has a very positive impact on the research and development of clinical medicine. In addition, the three-dimensional model of the reconstruction of organs and tissues according to the two-dimensional slice sequence is widely used in the field of medical imaging analysis. It can reconstruct the three-dimensional model of various organs, predict the tissue lesions and the human body. It can not only improve the accuracy of diagnosis, but also provide an omni-directional, three-dimensional organ model for doctors and relevant scholars, which is convenient for research and depth analysis. From the purpose of the study, this paper first introduced the DICOM standard, then converted medical images into bitmap images, and used it. The unrelated display mode displays the bitmap. According to the importance of the information, there is a selective reading of important information related to the patient, equipment and image parameters in the head of DICOM file. According to the low resolution and contrast of medical image, the image preprocessing method is designed in this paper. Gray scale transformation and image enhancement, and many kinds of geometric transformations commonly used in translation, and image enhancement is divided into smooth and sharp. Compared with traditional medical image segmentation methods, this paper proposes a method of automatic segmentation of multiple organs from abdominal medical image data. The purpose of this study is to construct super pixels and use them. More accurate segmentation is predicted. The super pixel based segmentation method ADAPs to various imaging conditions in the CT image and considers the relationship between multiple organs. The framework of this method is as follows: (1) clustering the super pixels according to the pixel correlation and location proximity (2) is introduced into the organ spatial structure distribution map, and the segmentation of multiple organs is modified. In our framework, we use the conventional single organ segmentation method to segment the accurate prediction organ, and use the super pixel clustering method to segment the residual organs. Further research on the segmentation results, study the Marching Cubes (MC) method on the basis of the intraperitoneal two-dimensional slice interested organ segmentation, the key point and the key point of the algorithm. The main difficulties are to determine the relationship between the voxel and the equivalent surface, and to find the normal vector of the equivalent surface at the time. Finally, the three-dimensional model of the specific organs in the abdominal cavity after the reconstruction is given.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】