【作者】 罗晓华；

【导师】 何中市；

【作者基本信息】 重庆大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 肺癌是世界上最常见的内脏恶性肿瘤之一,也是确诊后存活率最低的癌症之一。在我国城镇人口中,肺癌死亡率已居肿瘤死亡率首位。更重要的是,目前对肺癌的确诊又常常是它的中晚期,给病人的生命带来更大的威胁。如何对肺癌实现早期诊断,是人们、特别是病人和医生关注的首要问题。因此本文把肺癌的早期诊断作为研究的主要任务和目标。其中肺癌的早期阶段肺结节的特征提取是本文研究的主要方向,包括以下几个主要研究方面:1.结节的初分类和提取在分形理论的基础上,提出了用加权分数维对低对比度感兴趣区域进行灰度增强的方法,找到了作为自适应阈值判断准则的系数(分形维数),结合最大类间方差法(OTSU方法),再利用数学形态学进行边缘检测,提取了疑似的封闭感兴趣(ROI)。2.结节几何形态的分形特征提取分叶与毛刺在肺内孤立小结节(Solitary Pulmonary Nodules,SPN)的良恶性诊断中占有十分重要的地位,并采用形状分形和形状多重分形奇异谱方法,对结节分叶与毛刺进行形状分析,提取了良恶性结节的分形形状特征。3.结节纹理的分形特征提取计算结节内中心象素的方向分形(各向异性)、方向多重分形奇异谱方法,分析了结节的多重分形谱和分维数的特征,并结合区域的分维数及多重分形,实现了对结节形状和纹理分形特征的分析和提取,4.各种结节的具体病症的定量分析目前人们对结节的研究主要是对它的良恶性质的定性判别上,很少有人对具体病症进行分析、归类。未见对具体病症采用形状分形、方向分形和多重分形奇异谱对肺结节的形状、纹理等特征提取利用分形理论进行系统研究的,本文在这方面做了初步研究,并提出了一套对肺内孤立小结节进行定量分析的方法。结合结节的形态分形特征和纹理分形特征得到了可以初步判断其具体病症的有用信息,利用马氏距离进行分类判别,实现对肺癌的初步早期诊断。更多还原

【Abstract】 Lung cancer is the most common malignant tumor and one of the lowest livability tumors after diagnosis as is known so far. It is increasing annually and now the first cause of cancer-related mortality in cities. In order to improve the survival rate of lung cancer patients, detection in the early stages has a significantly more hopeful prognosis and is the key treatment.The dissertation focuses on extracted features of Shape of fractal and Anisotropic Fractal from nodules of the earlier period stage Lung cancer, mian aim is able to quantitative analysis the disease types of nodules.1. Nodules rough Classification Base one the fractal theory , propose the method of enhancement the low -contrast of ROI region and the coefficient as a criterion for discrimination is found based on a new kind of weighted differential box-counting algorithm. In orde to segmentation a region of ROI image by combined the OTSU method and morphological , then to be extracted SPN from pulmonary parenchyma.2. Shape fractal features extraction of noules Lobulation and Burr play a significant role in distinction between benign and malignant Solitary Pulmonary Nodules diagnosis. And benign and malignant lung nodules differ in lobulation and burr features. In this chapter, features of malignant lung nodules are analyzed based on shape of fractal theory.3. Texture features extraction of nodules Anisotropic Fractal is used to analyze and extract the texture features of lung nodules. This method can compute different dimensions in different directions.4. Quantitative analysis the disease types of nodules The experiment indicates that fractal features are effective on reflecting the texture and edge shape of lung nodules. Further more, more feature information can be gained from multifractal than Isotropic fractal. And this method can be applied to any texture and shape images. Then Mahalanobis discriminant analysis is used to classify nodules,the discriminant scores are analyzed using Shannnon entrop method.更多还原