基于HRCT影像组学、临床特征与SUV_(max)的联合模型列线图预测cT_(1)N_(0)M_(0)期肺腺癌脏层胸膜浸润的价值  

Value of a prediction model combined of radiomics,clinical features and SUV_(max) to visceral pleural invasion in cT_(1)N_(0)M_(0) lung adenocarcinoma

作　　者：张莲[1] 黄钢[2] 赵佳壁 孙珂 魏翔宇 孙希文[3] ZHANG Lian;HUANG Gang;ZHAO Jiabi;SUN Ke;WEI Xiangyu;SUN Xiwen(Department of Radiology,Shanghai Pulmonary Hospital Affiliated of Tongji University,Shanghai200433,China.)

机构地区：[1]上海市嘉定区中医医院放射科,上海201800 [2]上海健康医学院上海市分子影像学重点实验室,上海201318 [3]同济大学附属上海市肺科医院放射科,上海200433 [4]复旦大学附属华山医院放射科,上海200040 [5]上海中医药大学附属曙光医院针灸科,上海201203

出　　处：《中国中西医结合影像学杂志》2023年第3期239-246,共8页Chinese Imaging Journal of Integrated Traditional and Western Medicine

基　　金：上海市嘉定区农业和社会事业科研项目(JDKW-2018-W29);上海市科委科技创新计划项目(21Y11910400)。

摘　　要：目的:探讨基于HRCT图像的影像组学、临床特征与最大标准化摄取值(SUV_(max))的列线图预测cT_(1)N_(0)M_(0)期肺腺癌脏层胸膜浸润(VPI)的价值。方法:回顾性分析经手术病理证实为cT_(1)N_(0)M_(0)期肺腺癌,并行HRCT和18F-FDG PET/CT扫描的155例患者,共157个病灶,按7∶3的比例随机分为训练集(110个病灶)和测试集(47个病灶)。从CT图像中共提取1158个影像组学特征。应用影像组学分析计算影像组学评分,并结合临床特征、SUV_(max)和CT特征,构建预测胸膜浸润的联合模型,以列线图进行可视化。采用ROC曲线、校准曲线和决策曲线对模型性能进行评估。结果:联合模型的多因素逻辑回归分析结果显示,影像组学评分(OR=3.14,P=0.002)、CT观测结节与胸膜的关系(OR=2.79,P<0.001)、SUV_(max)(OR=1.31,P=0.035)为肺腺癌VPI的独立预测因子。联合预测模型在测试集的AUC值、准确率、敏感度、特异度分别为0.869、0.787、0.909、0.750,明显高于临床模型和单独的影像组学评分。结论:基于HRCT图像的影像组学、临床特征与SUV_(max)的联合模型列线图可准确预测cT_(1)N_(0)M_(0)期肺腺癌患者VPI情况,有望成为临床术前评估肺腺癌VPI的可靠工具。Objective:To establish a nomogram model based on radiomics,clinical features and SUV_(max) for preoperative prediction of visceral pleural invasion(VPI)in patients with cT_(1)N_(0)M_(0) lung adenocarcinoma.Methods:A retrospective analysis of 155 patients with surgically pathologically confirmed cT_(1)N_(0)M_(0) lung adenocarcinoma who underwent HRCT and 18F-FDG PET/CT scans was performed.157 lesions were found and randomly assigned to a training cohort(110 lesions)and a validation cohort(47 lesions)at a ratio of 7∶3.A total of 1158 radiomics features were extracted from CT images.Radiomics analysis was used to calculate the radiomics score(rad-score),and the combined model for predicting VPI was constructed by combining the clinical features,SUV_(max) and rad-score,and visualized by a nomogram.The model performance was evaluated by ROC curve,calibration curve and decision curve analysis(DCA).Results:The results of multiple logistic regression analysis of the combined model showed that the rad-score(OR=3.14,P=0.002),the relationship between nodule and pleura showed on CT images(OR=2.79,P<0.001)and SUV_(max)(OR=1.31,P=0.035)were the independent risk factors for predicting VPI.The AUC,accuracy,sensitivity and specificity of the combined model in the validation cohort were 0.869,0.787,0.909 and 0.750,respectively,which were significantly higher than those of the clinical model and the rad-score.Conclusions:The combined model based on radiomics,clinical features and SUV_(max) visualized by a nomogram can accurately predict VPI in patients with cT_(1)N_(0)M_(0) lung adenocarcinoma.It is expected to be a reliable tool for the preoperative evaluation of VPI.

关 键 词：肺肿瘤 腺癌 体层摄影术 X线计算机 正电子发射断层显像术 最大标准化摄取值 脏层胸膜浸润 列线图 

分 类 号：R734.2[医药卫生—肿瘤]