CT平扫图像纹理分析对预测肺癌纵隔淋巴结转移的价值研究  

Unenhanced CT Image Texture Analysis in Predicting Mediastinal Lymph Node Metastasis of Lung Cancer

作　　者：曹恩涛 陈颖[1] 蔡庆[1] 陈双庆[1] 刘晨鹭[1] 张帆 CAO Entao;CHEN Ying;CAI Qing(Department of Radiology,Suzhou Hospital Affiliated to Nanjing Medical University,Suzhou,Jiangsu Province 215001,P.R.China)

机构地区：[1]南京医科大学附属苏州医院放射科,苏州215001

出　　处：《临床放射学杂志》2022年第12期2219-2223,共5页Journal of Clinical Radiology

基　　金：国家自然基金资助项目(编号:81871353)。

摘　　要：目的评价CT平扫图像纹理分析在肺癌患者良性与转移性纵隔淋巴结鉴别诊断中的价值。方法回顾性分析74例经手术证实的原发性肺恶性肿瘤的术前CT平扫图像,从114个纵隔淋巴结的CT平扫图像中提取不同纹理特征参数,进行良恶性组间统计分析。多个纹理参数及淋巴结大小定量参数组合进行二元Logistics回归分析、受试者工作特征(ROC)曲线分析,并计算曲线下面积(AUC)。结果良恶性组间8个图像纹理特征存在差异,图像纹理特征模型AUC为0.778(P<0.001);纹理特征联合短径预测模型AUC为0.780(P<0.001);纹理特征联合横截面积预测模型AUC为0.813(P<0.001),最佳临界值敏感度和特异度分别为71.7%和78.7%。结论CT平扫图像纹理分析有助于肺癌纵隔淋巴结的术前良恶性判断。Objective To evaluate the value of CT image texture analysis in the differential diagnosis of benign and malignant mediastinal nodules in patients of lung cancer.Methods Images of unenhanced CT scans before operation,from 74 patients with primary pulmonary malignant tumor confirmed by operation,were studied retrospectively.Different texture characteristic parameters were extracted from 114 lymph node unenhanced CT images and analyzed statistically between benign and malignant groups.Multiple texture parameters and size quantitative parameters of lymph nodes were combined for binary Logistics regression analysis,Receiver operating characteristic(ROC)analysis and area under curve(AUC)calculation.Results 8 texture features were different between benign and malignant groups,and the area under ROC curve of the image texture model was 0.778(P<0.001).The area under ROC curve of the prediction model based on texture feature combining short axis diameter was 0.780(P<0.001).The area under ROC curve of the prediction model based on texture feature combining cross-sectional area was 0.813(P<0.001),Using optimum-threshold criteria,the diagnostic sensitivity and specificity were 71.7%and 78.7%,respectively.Conclusion Unenhanced CT image texture analysis can help differentiate mediastinal lymph nodes from benign and malignant in patients before operation with lung cancer.

关 键 词：肺癌 体层摄影术 X线计算机 淋巴结 纹理分析 

分 类 号：R734.2[医药卫生—肿瘤] R730.44[医药卫生—临床医学]