分形理论和图像纹理特征在肺结节诊断中的应用

The application of fractal theory and imaging textural feature for diagnosing pulmonary nodule

目的:探讨分析理论和图像纹理特征分析法在肺结节诊断中的作用。方法:收集104例经临床手术病理证实的肺结节患者,采用水平集模型自动分割、提取肺结节,用分维数和灰度共生矩阵法对良、恶性结节特征进行比较。结果:运用不同大小的"盒"计算出的分维数(DF)值不同:肺癌结节DF1=1.82±0.140,DF2=1.78±0.137,DF3=1.70±0.138,DF4=1.64±0.140。非肺癌结节:DFI=1.74±0.144,DF2=1.64±0.201,DF3=1.54±0.227,DF4=1.50±0.207;肺良性结节的能量、对比度、逆差距、相关性及熵分别为0.984 15±0.014 59,0.019 53±0.010 56,0.999 65±0.000 19,0.973 79±0.011 29,0.061 76±0.048 03;肺恶性结节的能量、对比度、逆差距、相关性及熵分别为0.972 23±0.022 15,0.026 29±0.011 62,0.999 53±0.000 20,0.958 40±0.034 88,0.099 89±0.065 19。结论:肺良恶性结节图像纹理及分维数存在差异,基于灰度共生矩阵图像纹理统计方法和分形理论对肺结节的CT诊断有重要的辅助价值。

Objective：To prospectively study The application of fractal theory and imaging textural feature for diagnosing pulmonary nodule.Methods：104 cases which were confirmed by clinical surgery pathology of patients with pulmonary nodules were collected.The cases were analyzed by the level set model automatic segmentation and extraction of pulmonary nodules,compared benign and malignant nodule feature by fractal dimension and gray level co-occurrence matrix characteristics.Results：using different sizes of ＂box＂ to calculate the fractal dimension（DF） value of the different：The lung cancers nodules DF1=1.82 and 0.140,DF2=1.78±0.137,DF3=1.70±0.138,DF4=1.64±0.140.Not lung cancers nodules：DF1）=1.74±0.144,DF2=1.64±0.201,DF3=1.54±0.227,DF4=1.50±0.207;Energy,contrast,distance,correlation and entropy of benign nodule were 0.984 15±0.014 59,0.019 53±0.010 56,0.999 65±0.000 19,0.973 79±0.011 29,0.061 76±0.048 03,and energy,contrast,distance,correlation and entropy of malignant nodule were 0.972 23±0.022 15,0.026 29 ±0.011 62,0.999 53±0.000 20,0.958 40±0.034 88,0.099 89±0.065 19.Conclusion：The image texture feature and fractal dimension of benign and malignant lung nodules were difference.The method which based on the gray level co-occurrence matrix of image statistics and fractal theory to CT diagnosing pulmonary nodule was significant.