应用Sandwich Logistic回归分析探讨肺小结节恶性病变影响因素  

Exploring the Risk Factors of Lung Pulmonary Nodules with Cancer Using Sandwich Logistic Regression Analysis

作　　者：郭秀花[1] 张勇[1] 王瓛[1] 李坤成[2] 姚新宇[2] 梁志刚[2] 

机构地区：[1]首都医科大学公共卫生与家庭医学学院,北京100069 [2]首都医科大学宣武医院

出　　处：《数理医药学杂志》2008年第1期66-68,共3页Journal of Mathematical Medicine

基　　金：北京市教育委员会科技发展计划面上基金课题(编号:KM200610025014)资助;首都医科大学基础与临床基金课题(2006JL57)资助

摘　　要：目的:探讨肺小结节良恶性鉴别诊断的定量分析模型。方法:对21名3208张肺小结节良恶性肺部CT图像,利用灰度共生矩阵法,借助MATLAB和VC++软件平台提取纹理特征,利用SAS9.1版本统计学软件进行Sandwich Logistic回归分析。结果:入选的纹理特征有能量值、熵值、惯性炬、局部平稳性和灰度值,系数检验的P值均为<0.0001;且能量值、熵值和惯性炬值越小时,局部平稳性和灰度值越大时,越可能是恶性肺小结节。结论:提取纹理特征对于区分图像的良恶性有较强的描述能力。Objective： To explore the risk factors of lung pulmonary nodules with cancer. Method： Statistical analysis were performed using MATLAB,VC ＋ ＋ to extract the textures of the 3208 lung CT images. Then the collected data was analyzed by SAS 9.1 using sandwich logistic regression analysis. Result. There are significant difference in energy, entropy, inertia inverse difference moment and gray value between the normal lung CT images and lung CT images with tumor. Conclusion： Numerical-contrast moments have a good description ability in distinguishing images.

关 键 词：肺部患者 CT图像 纹理提取 肺小结节 

分 类 号：R311[医药卫生—基础医学]