血管三维辅助诊断系统的研究

发布时间：2018-02-26 11:17

本文关键词： 血管图像三维重建血管分割中轴线提取　出处：《四川师范大学》2016年硕士论文　论文类型：学位论文

【摘要】：近些年,随着大量的医学辅助手段如CT、MRI在临床诊断的应用,对心脑血管疾病的诊断状况已经有所改善。这些诊断方式只是以二维切片的方式显示人体组织和器官,诊断时常常依赖较为丰富的经验。对于经验尚浅的医生来说,这种方式并不直观。本文旨在通过对血管造影断层图像实现三维重建并提取血管三维模型的中轴线,为医生在血管病症方面诊断提供直接可靠的信息。本文根据血管三辅助维诊断系统的需求,通过实验对比目前较为成熟的预处理技术,为下一步工作提供高质量图像。对表面重建和体重建的三维重建算法进行了分析,借助可视化开源库VTK、MITK实现切片集重建。研究了图像分割技术的各种算法,提出了一种改进的二维最大交叉熵分割方法。该方法通过对切片直方图统计的高斯分布进行随机采样,减小原始算法对前景和背景先验分布的误差。替代了直接用高斯分布计算目标和背景的概率,得出更加合理的阈值向量。从二维距离变换提取骨架线方法入手,提出了一种基于动态规划与区域搜索思想的二维距离变换计算方法。和传统距离变换方法进行对比,分析了算法的利弊。并将这种方法运用到三维距离变换中,提取三维血管中轴线。在获取“粗”中轴后,针对中轴线存在像素堆积或锯齿这样的不足,通过曲线光顺性去除冗余像素,利用插值算法对中轴进行优化。最后,从软件工程的角度讨论了本文研究内容基础上实现的一个简易三维可视化辅助诊断系统。本文提及和实现内容虽无法满足真正实践的需求,却可以从中展望到未来数字图像处理技术和三维可视化技术在医学领域中的重要作用,这必然会成为医学诊断的趋势。
[Abstract]:In recent years, the diagnosis of cardiovascular and cerebrovascular diseases has been improved with the application of a large number of medical aids such as CTN MRI in clinical diagnosis. The diagnosis often depends on rich experience. This method is not intuitionistic for the doctors who have little experience. The purpose of this paper is to reconstruct 3D angiography images and extract the central axis of 3D vascular models. To provide direct and reliable information for doctors in the diagnosis of vascular diseases. According to the needs of vascular three-dimension diagnostic system, this paper compares the current mature preprocessing technology through experiments. The 3D reconstruction algorithms of surface reconstruction and volume reconstruction are analyzed, and the slice set reconstruction is realized with the help of VTK MITK, a visual open source library. Various algorithms of image segmentation technology are studied. An improved two-dimensional maximum cross-entropy segmentation method is proposed, in which the Gao Si distribution of slice histogram is sampled randomly. The original algorithm reduces the error of foreground and background prior distribution. Instead of directly calculating the probability of target and background by Gao Si distribution, a more reasonable threshold vector is obtained. In this paper, a two-dimensional distance transform calculation method based on dynamic programming and region search is proposed. Compared with the traditional distance transformation method, the advantages and disadvantages of the algorithm are analyzed, and the method is applied to 3D distance transformation. After obtaining the "coarse" axis, the center axis is optimized by using interpolation algorithm to eliminate the redundant pixels through curve smoothing, aiming at the deficiency of pixel accumulation or sawtooth in the center axis. From the point of view of software engineering, this paper discusses a simple 3D visualization aided diagnosis system based on the research contents of this paper. Although the contents mentioned and implemented in this paper can not meet the needs of real practice, However, the important role of digital image processing technology and 3D visualization technology in medical field in the future can be forecasted, which will inevitably become the trend of medical diagnosis.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R54;TP391.41

【参考文献】