基于分形理论的肺部CT图像纹理分割算法研究

【作者】 孙涛；

【导师】 蒋爱平；

【作者基本信息】 黑龙江大学 ， 电子与通信工程， 2009， 硕士

【摘要】 近年来,空气污染严重,肺癌的发病率越来越高,临床上通过CT成像技术辅助诊断肺癌,在很大程度上仅凭医生肉眼进行定性和经验性判断,具有一定的局限性,且误诊率较高。本文旨在通过计算机对肺部软组织CT图像的纹理分析研究,进一步挖掘图像信息,为CT诊断肺癌提供一种计算机辅助诊断的客观手段。本文以分形理论为基础,通过分析多种描述纹理的方法及特点,研究并提出了一种肺部CT图像纹理分割算法,实现了肺部正常组织和异常的分割。该算法以分形维数作为纹理特征参量,利用分形布朗运动模型提取Hurst指数H,再利用H分布的累积直方图进行纹理分割。软件部分是由Matlab实现的,实验结果表明：该算法与现有的基于边缘的Canny分割算法和基于灰度分布的累积直方图分割方法相比,能更好地识别肿瘤区域。更多还原

【Abstract】 Recently, with air pollution being serious, incidence of lung cancer is now increasing. The diagnosis using CT mainly depend on the experiences of physicians, so there are often wrong diagnosis. The objective of this research is to provide a computer-aided method for the diagnosis of lung cancer by applying texture analysis theory into CT lung images’ analysis.After analyzing kinds of approaches to characterize texture, the thesis studies and presents a suite of system to recognize Cancer Structure in CT lung images based on fractal geometry. The system uses fractal Brownian motion to extraction Hurst parameter, and then applies histogram accumulated by H to segment texture. The software program is finished through Matlab. The experimental result indicates:the approach can complete Cancer Structure recognition better than that based on Canny and histogram accumulated by grey level.更多还原