《四川大学硕士学位论文.doc》由会员分享，可在线阅读，更多相关《四川大学硕士学位论文.doc（5页珍藏版）》请在三一文库上搜索。
1、四川大学硕士学位论文四 川 大 学硕 士 学 位 论 文题 目 基于 Power mode图像和骨架算法的血管直径测量方法作 者 李硕 完成日期 2009 年 04 月 01 日培 养 单 位 四川大学指 导 教 师 刘东权 教授专 业 计算机应用研 究 方 向 计算机应用授予学位日期 年 月 日1四川大学硕士学位论文基于 Power mode图像和骨架算法的血管直径测量方法计算机应用研究生： 李硕 指导教师： 刘东权在临床医学领域里，很多情况下医生希望测量病人血管的直径。它们对于很多病情的诊断，如冠心病、肝硬化等，起到了至关重要的作用。随着计算机硬件技术的飞速发展，我们可以利用图像处理技术更
2、方便地利用血管图像来帮助医生测量血管直径。进年来发展出的 3D 超声技术可以提供给我们更多的血管信息。 Power mode图像相比传统的 B 模式超声图像和多普勒彩色血流图（ CFM ），高信噪比的优点。而且它还具有可选观察角度，易于从背景中分割出来等特性，因此非常适合作为血管测量用图。本文给出了一个基于 3D Power mode 图像计算血管直径的方案：先将 3DPower mode 图像转化成二值图像，然后提取二值图像中血管的边界和中线，然后对中线上任意一点找到其切线，最后找到切线过这点的垂线，垂线与边界相交于两点之间的距离记为此中线点处的直径。整个方案框架中如何找到一条连续，准确，理
3、想的中线，是重点与难点，它决定了最终测量结果的准确性。而且由于血管形状复杂，边界不光滑，又有很多分支，就更加大了中线提取的难度。所以作者对此进行了深入的研究。在计算机视觉技术里有一类算法，是用来提取数字图像中某个或某些物体的中轴线，以便用来表征物体的拓扑结构，或作为物体识别的特征。我们称这类算法为骨架算法。有些文献也将其称为中轴变换，或细化算法。已经有大量的文献，基于不同思想，对骨架算法进行了各种实现和改进。这些方法的运算复杂度，得出的结果以及适用的情形都不尽相同。最常见的细化法，利用一些判断条件不断删除目标物体边界点，最后剩下骨架点，适用于成条状，带状物体的骨架提取，如文字，指纹；基于数学形
4、态学的开操作和腐2四川大学硕士学位论文蚀操作“沉淀出”目标物体骨架的算法，适用于寻找物体特征点，并可以从骨架重建出原物体来；基于目标物体欧式距离变换图提取出的骨架，拓扑结构较准确；利用几何里物体最大内切圆的概念寻找骨架点的方法，运算简单，速度快；等等。本文第一部分首先提出科研问题和科研背景，然后提出了一套解决方案，并概要地介绍了方案中用到的两个核心技术， Power mode 图像和骨架算法。第二部分对骨架算法进行了深入地研究，并于最后提出了一个基于欧式距离图，距离图梯度信息和最短路径思想的算法。最后一个部分介绍了作者在进行研究过程中看到的骨架算法在其他很多领域里的重要应用。简要地整理介绍出来
5、，希望能对从事相关工作的读者有所帮助。关键词： 血管直径测量 骨架算法 形态学 细化 距离变换 计算机视觉 3DPower mode3四川大学硕士学位论文Blood Vessel Tracking and Diameter MeasurementMethod Based on Skeleton AlgorithmMajor ：Computer ApplicationGraduate：Li Shuo Advisor：Dongc LiuIn clinic medicine field, doctors tend to know patientsblood vessel diameter in ma
6、nysituations, since they are very helpful to the diagnoses of many diseases, such as coronaryheart disease, hepatocirrhosis and so on. With the fast development of computer hardware,we can take use of digital image processing techniques to help doctors to measure the bloodvessel diameter convenientl
7、y.3D ultrasound techniques, which developed recently, can provide us much informationabout blood vessel. 3D Power mode image has much higher signal-to-noise ratio (SNR)comparing to traditional B mode ultrasound image and Doppler color flow map (CFM).Also it is angle dependence.What is more, blood fl
8、ow can easily be extracted frombackground in 3D power mode image, so it is a very good choice for blood vessel diametermeasurement.This paper presents a solution to calculate vessesl diameter based on 3D power modeimages: first transfer 3D Power mode image into binary image, then extract the binaryi
9、mages boundary and midline; after that, for each point on the midline, we find the tangentand the Perpendicular Line of the tangent through that midline point, the perpendicular line and theboundary intersect into two points, the distance between those two points are the diameter of the vesselon tha
10、t given midline point.In this solution, the key is to find a continuous, exact, and one pixel wide midline, since it willaffect the veracity of the final measurement. However, the structures of the vessels are usually verycomplicate, and the boundary is not smooth, which increase the difficulty of e
11、xtracting the midline. So4四川大学硕士学位论文we will pay more attention onthis part.In computer vision field, there is a class of algorithms which are used to extractobjects midline from the digit image, so that objects topology structure can be wellexpressed or objects feature can be better found out. They
12、are called skeleton algorithm, orin some other literature, thinning or midline transform.Based on different ideas and situations, thousands of papers have presented manykinds of achievements and improvements on skeleton algorithm, during which, thecomplexity of calculation, getting results and suita
13、ble situation are different. The mostfrequently used method, thinning, which takes use of some boundary condition judgmentsto delete boundary points layer by layer and leaves the skeleton points in the end, is moresuitable for the extraction of strip objects skeleton, such as handwriting, and finger
14、prThe methods that based on mathematics morphology ospen and erosion operation, aremore suitable for looking for objects feature points, and also can be used to trace objectsback from their skeleton points. Methods based on Euclidean distance map of objects canget skeletons with more accurate geomet
15、ry topology. Methods that take use of objectsmaximum inside tangent circles are much easier to calculate, and usually run fast.The first part of the article is an introduction of the research topic and researchbackground, the solution structure of the problem and two core techniques used for theprob
16、lem solving: Power mode image and skeleton algorithm.In the second part, we did some deep research on the skeleton algorithm. And alsopresent a skeleton algorithm based on Euclidean Distance Map, gradient information of thatmap and the shortest path algorithm. The results show that it is very good t
17、o representvessels structure.In the last part of the article, we will introduce the application of skeleton algorithm inother fields. Hopefully it is helpful to readers in other fields.Key words: Measurement of Vessel Diameter, Skeleton Algorithm, Thinning,Mathematics Morphology, Distance Transform, 3D Power mode.5