【作者】 王健；

【导师】 李海峰；

【作者基本信息】 哈尔滨工业大学 ， 计算机科学与技术， 2009， 硕士

【摘要】 医学图像三维重建,是计算机图形学和数字图像处理技术在生物医学工程中的重要应用。医学三维重建技术已经成为医院诊断、治疗的重要依据和手段,在医学诊断、手术规划、模拟仿真、整形及假肢外科、解剖教学等方面都有重要的应用。本文对互相缠绕,结构复杂的血管进行三维重建。先用面绘制方法中的移动立方体经典算法进行了重建,针对管状结构,本文又提出了一种自动提取三维中心线并且估计其半径来描绘血管物体的方法。首先,面绘制方法存在不能反映整个原始数据场的全貌及细节的问题,本文采用提取三维中心线并且估计其半径来描绘血管物体。本文先对切片图像进行预处理,以得到物体间对比最大化,然后用带标记控制的分水岭算法分割互相缠绕的血管。在最优路径上的有最小花费的点定义为前景标记,分割区域的质心用来调整搜索到的中心线点。这种方法可以快速准确的提取各种情况的中心线,处理诸如横渡截面,重合分割面的复杂血管结构。其次,提取血管的三维中心线。提取时,包括两个步骤:先用三维线性过滤方法为每一个点分配一个相似度测量,然后在每一个点内搜索区域本文寻求一个最优路径。再次,针对文献中的方法估计血管不准的问题,我们在估计半径时,先对分割后的区域进行裁剪,然后对裁剪过分的部分进行合并,最后对半径进行估计。这样估计出的半径能够全部覆盖血管的主要信息,准确的表达出血管结构。本文在描绘立体图时用椭圆来描绘血管的形状。最后,针对现有方法半自动需要手工操纵的缺点,本文提出一种全自动的方法。更多还原

【Abstract】 Three-dimensional reconstruction of medical images is a multi-disciplinary subject. It is an important application of computer graphics and image processing in biomedicine engineering, which has become a helpful means for clinic diagnosis and treatments. Three-dimensional reconstruction and visualization of medical images are widely used in diagnostic, surgery planning and simulating, plastic and artificial limb surgery, and teaching in anatomy.This paper focuses on three-dimensional reconstruction of complicated vessels structures such as cross-over sections and attaching segments. The principle and implementation of the Marching Cubes, a well-known surface extraction algorithm, are demonstrated, so we propose an automated three-dimensional centerline extraction and vessel radius estimation for three-dimensional reconstruction tool to assist in this task.First, due to Marching Cubes cannot reflect the raw data, we propose a three dimensional centerline extraction and vessel radius estimation for three dimensional reconstruction tool. The contrast of the objects is maximized by preprocessing, and an marker-controlled watershed algorithm is used to segment different vessel objects in order to distinguish axons that are attaching together. The points on the optimal paths with the minimum cost values are regarded as the foreground markers. The centroids of the segmented regions are used to adjust the searched centerline points using DP. The proposed method can rapidly and accurately extract multiple vessel centerlines and can handle.Second, we propose a highly automated three-dimensional centerline extraction tool to assist in this task. It consists of two steps. In the first step, every point in the image stack is assigned a similarity measurement to the three-dimensional line structure using the three-dimensional line filtering method. In the second step, we search for an optimal path for each point within the search region on the current slice to the centerline points on the previous slice.Third, because the inaccuracy of estimating the radius of the vessels in the literature method, we cut and merge the regions after segmentation to estimate the radius of the vessels for three-dimensional vessel reconstruction. They cover all the key messages and accurate expression of the vessels structures. And then we describe the blood vessels with an oval shape.At last, because the method is semiautomatic, we design a fully automated algorithm.更多还原