CT图像放射组学分析鉴别诊断磨玻璃结节肺腺癌的浸润性  

Radiomic features in CT to differentiate invasive pulmonary adenocarcinomas from non-invasive pulmonary adenocarcinomas with part-solid ground-glass nodules

作　　者：赵志雄[1] 肖运平[1] 刘铁军[1] ZHAO Zhixiong;XIAO Yunping;LIU Tiejun(Department of Radiology,Liuzhou General Hospital,Liuzhou 545006,China)

机构地区：[1]柳州市人民医院放射科,广西柳州545006

出　　处：《实用放射学杂志》2021年第9期1441-1445,1479,共6页Journal of Practical Radiology

摘　　要：目的通过CT图像放射组学分析鉴别诊断磨玻璃结节(GGN)肺腺癌的浸润性,为侵袭性肺腺癌(IPA)制定诊断诺模图模型.方法回顾性选取经手术确诊的88例患者,共100个亚实性结节.选取增强CT动脉期图像进行三维结节感兴趣区(ROI)的分割并计算定量放射组学特征.使用逻辑回归分析将一组常规临床风险因素和放射医生视觉评估的定性CT特征与放射学特征进行比较.建立3种诊断模型,即使用临床风险因素和CT定性特征的基础模型,使用包含具有统计学意义的放射组学特征模型,以及结合所有重要特征的诺模图模型,并根据受试者工作特征(ROC)曲线对3种模型的诊断效能分别进行比较.结果除了3个视觉评估的CT定性成像特征外,还发现从数百个放射学特征中选择的另外3个定量特征与诊断IPA显著相关(P<0.05).用诊断诺模图模型区分IPA与非IPA的性能最佳[曲线下面积(AUC)=0.903],均高于基础模型(AUC=0.853,P=0.0009)或放射组学模型(AUC=0.769,P<0.0001).决策曲线分析也表明在临床诊断中使用此诺模图模型的潜在益处.结论由CT图像放射组学和基本特征整合构建的诺模图模型,对IPA具有更优良的鉴别诊断性能.Objective To analyze and diagnose the invasiveness of ground-glass nodules(GGN)lung adenocarcinoma through CT image radiomics,and to develop a diagnostic nomogram model for invasive pulmonary adenocarcinomas(IPA).Methods Data of 88 patients with 100 sub solid nodules confirmed by surgery were collected retrospectively.Arterial-phase contrast-enhanced CT images were used to 3D nodules image segmentation.Quantitative radiomic features were extracted automatically by models.A set of regular clinical risk factors and visually-assessed qualitative CT imaging features were compared with the radiomic features respectively using Logistic regression analysis.Three diagnostic models including a basis model using the clinical factors and qualitative CT features,a radiomics model using significant radiomic features,and a nomogram model combining all significant features were established.Diagnostic powers of three models were compared by receiver operating characteristic(ROC)curves.Results In addition to three visually-assessed qualitative imaging features(solid component size,solid components proportion,and pleural indentation),another three quantitative features(ClusterShade_angle45_offset7,sum Entropy,and Spherical Disproportion)selected from hundreds of radiomic features were found to be significantly associated with IPA(P<0.05).The diagnostic nomogram model showed a significantly better performance[area under the curve(AUC)=0.903]in differentiating IPA from non-IPA than both the basis model(AUC=0.853,P=0.0009)and the radiomics model(AUC=0.769,P<0.0001).The potential benefit of using such a nomogram model in clinical diagnosis was suggested by decision curve analysis.Conclusion The nomogram model constructed by the integration of CT image radiomics and basic features has better differential diagnosis performance for IPA.

关 键 词：放射组学 计算机体层成像 肺腺癌 磨玻璃结节 

分 类 号：R814.42[医药卫生—影像医学与核医学] R734.2[医药卫生—放射医学]