基于CT影像的肺结节辅助诊断系统的设计与实现

发布时间：2019-03-07 11:52

【摘要】：癌症如今已成人类死亡的首要原因之一，而肺癌以高发病率，高死亡率成为当今最致命的癌症。尽早发现肺癌是治疗肺癌的最好方法。CT是检查肺癌的最佳手段之一，而精度越来越高的CT技术在得到更清晰图像的同时，数据量也大大增加，医生已不堪重负。速度快，不知疲倦的计算机成为帮助医生诊断的有效工具。本文对肺结节检测的相关技术进行研究，设计并实现了基于CT影像的肺结节辅助诊断系统。由于良性阴影和恶性阴影都有不同的种类，大小形态多样，训练样本仅仅分为良性和恶性两大类使各个类内相似程度低。为解决这个问题，，以及众多分类器利用欧氏距离度量方面存在的缺陷，本文提了出一种结合模糊C均值（Fuzzy-c means，FCM）聚类算法和马氏距离的分类器。通过对已有样本进一步细分，增加各小类样本的相似程度，然后用马氏距离代替欧氏距离的方法，计算样本与小类整体相似度，以提高分类的准确性。本文对ITK和VTK工具包进行研究，结合MFC搭建系统平台。通过对医生诊断肺结节的过程进行研究，将整个系统分为图像读取，肺部区域分割，肺部疑似阴影检测，疑似阴影的判别，图像显示和图像标记及交互七个模块。通过对各个模块应用技术的研究，设计并实现了相应功能。系统使用ITK和VTK工具包实现了图像的读取，显示和图像标记及交互功能；利用最优阈值法和区域连通算法进行肺部区域粗分割，再用空洞填充算法和修补算法完成肺部区域的修补；应用可变环形滤波器对肺部疑似阴影进行检测；最后利用本文提出的分类器进行良性和恶性的划分。通过对真实患者的图像进行检测，本系统有较好的敏感性。
[Abstract]:Cancer is now one of the leading causes of human death, and lung cancer is the deadliest cancer of the day with high morbidity and mortality. Early detection of lung cancer is the best way to treat lung cancer. Ct is one of the best methods to examine lung cancer. While more and more accurate CT technology can get clearer images, the amount of data increases greatly, and doctors are overburdened. Fast, indefatigable computers have become an effective tool to help doctors diagnose. In this paper, the correlative technology of pulmonary nodule detection is studied, and the assistant diagnosis system based on CT image is designed and realized. Because there are different kinds of benign shadows and malignant shadows, the training samples are divided into two categories: benign and malignant, so the degree of similarity in each category is low. In order to solve this problem and the shortcomings of many classifiers in Euclidean distance measurement, a classifier combining fuzzy C-means (Fuzzy-c means,FCM) clustering algorithm and Mahalanobis distance is proposed in this paper. In order to improve the accuracy of classification, the similarity degree of each sample is increased by further subdivision of the existing samples, and then the Mahalanobis distance is used instead of the Euclidean distance to calculate the global similarity between the samples and the small classes. In this paper, ITK and VTK toolkit is studied, and the system platform is built with MFC. By studying the process of diagnosing pulmonary nodules, the whole system is divided into seven modules: image reading, lung region segmentation, lung suspected shadow detection, suspected shadow discrimination, image display and image marking and interaction. Through the research on the application technology of each module, the corresponding functions are designed and realized. The system uses ITK and VTK toolkit to realize the functions of image reading, display, image marking and interaction. The optimal threshold method and the region-connected algorithm are used to segment the lung region rough, then the cavity filling algorithm and the patch algorithm are used to complete the lung region repair, and the variable ring filter is used to detect the suspected shadow of the lung. Finally, the classifier proposed in this paper is used to classify benign and malignant. By detecting the images of real patients, the system has a good sensitivity.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R734.2;R816.41;TP391.41

【参考文献】