CT图像纹理分析鉴别诊断磨玻璃密度肺腺癌的浸润性

发布时间：2019-01-07 12:38

【摘要】：目的探讨CT图像纹理分析鉴别诊断表现为磨玻璃密度结节的肺腺癌浸润性的价值。方法收集在我院接受肺部CT检查且手术病理证实为肺腺癌患者100例(浸润性腺癌56例,非浸润性腺癌44例)。随机选择69例为训练组,31例为验证组。使用A.K.(Analysis-Kinetics)分析软件进行影像特征提取;Kruskal-Wallis非参数检验和Spearman相关性分析进行特征降维;使用R语言软件包"GLM"函数,建立Logistic回归模型;以交叉验证方法对回归模型进行检验。采用ROC曲线评价独立预测因素的诊断效能。结果影像特征提取得到396个影像组学特征,经降维最终得到与鉴别肺非浸润腺癌与浸润腺癌最相关的参数3个,建模后验证Logistic回归模型示其诊断准确率为83.30%,敏感度及特异度分别为77.80%、91.70%。结论 CT图像纹理分析可有效鉴别表现为磨玻璃密度结节肺腺癌的浸润性。
[Abstract]:Objective to evaluate the value of CT texture analysis in differential diagnosis of lung adenocarcinoma with ground-glass density nodules. Methods 100 cases of lung adenocarcinoma (56 cases of invasive adenocarcinoma and 44 cases of non-invasive adenocarcinoma) were examined by CT in our hospital and proved by pathology. 69 cases were randomly selected as training group and 31 cases as validation group. Using A.K. (Analysis-Kinetics) analysis software for image feature extraction, Kruskal-Wallis nonparametric test and Spearman correlation analysis for feature dimensionality reduction, using R language package "GLM" function to establish Logistic regression model; The regression model was tested by cross-validation method. The ROC curve was used to evaluate the diagnostic effectiveness of independent predictive factors. Results three hundred and ninety-six imaging features were obtained by feature extraction. The most relevant parameters for differentiating noninvasive and infiltrating adenocarcinoma of lung were obtained by dimensionality reduction. The diagnostic accuracy of Logistic regression model was 83.30% after modeling. The sensitivity and specificity were 77.80 and 91.70, respectively. Conclusion texture analysis of CT images can effectively distinguish the infiltrative features of lung adenocarcinoma with ground-glass density nodules.
【作者单位】：中国医科大学附属第一医院放射科;GE医疗;
【基金】：公益性行业科研专项(201402013) 国家自然科学基金青年科学基金(81301222) 辽宁省科技厅项目(2012020073-302)
【分类号】：R730.44;R734.2

【相似文献】