CT影像组学模型预测非小细胞肺癌骨转移的价值  

Value of CT radiomics model in predicting bone metastasis of non-small cell lung cancer(NSCLC)

作　　者：夏熙婷 秦浩 张业林 许晓燕[1] 周永[1] XIA Xiting;QIN Hao;ZHANG Yelin;XU Xiaoyan;ZHOU Yong(Diagnostic Imaging Center,the Affiliated Tumor Hospital of Xinjiang Medical University,Urumqi 830011,China;Grade 2023 of Clinical Medicine,Xinjiang Medical University,Urumqi 830017,China)

机构地区：[1]新疆医科大学附属肿瘤医院影像诊断中心,乌鲁木齐830011 [2]新疆医科大学临床医学部,乌鲁木齐830017

出　　处：《新疆医科大学学报》2025年第2期162-169,共8页Journal of Xinjiang Medical University

基　　金：新疆维吾尔自治区产学合作协同育人项目(2023210016)。

摘　　要：目的探讨CT影像组学模型对非小细胞肺癌(NSCLC)骨转移的预测价值。方法回顾性分析2014年12月至2020年12月在新疆医科大学附属肿瘤医院经穿刺活检或手术病理证实诊断为NSCLC的149例患者的临床及影像资料。依据检查结果及临床分期信息,将患者分为骨转移组71例和非骨转移组78例。采用完全随机的方法将所有患者按6∶4的比例分为训练集(89例)和验证集(60例),应用ITK-SNAP软件,分别在平扫、动脉期和静脉期图像上逐层手动勾画病灶的感兴趣区(ROI)并进行三维融合获得病灶的容积ROI(VOI),然后导入AK软件中提取每个病灶的纹理特征。采用最小冗余最大相关(mRMR)和最小绝对值收敛和选择算子(LASSO)分别对平扫、动脉期、静脉期及多序列(平扫+动脉期+静脉期)的组学特征进行筛选用于骨转移的预测,采用Logistic回归分析建立影像组学模型并计算每例患者的分值(rad-score),采用100次留P交叉验证法判定其可靠性。通过Logistic回归分析将临床和影像资料中组间差异有统计学意义的变量建立常规模型,并与预测效能最高的影像组学模型联合建立综合诊断模型,并绘制其诺模图来分析预测概率。采用受试者工作特征(ROC)曲线评价模型的预测能力,应用决策曲线分析(DCA)评估模型的临床应用价值。结果本研究建立的多序列影像组学模型在训练集和验证集中预测NSCLC伴骨转移的曲线下面积(AUC)分别为0.858和0.848,优于常规模型(训练集和验证集的AUC分别为0.508和0.496)。肿瘤毛刺征与NSCLC患者肺癌伴骨转移密切相关(训练集OR=3.467,P=0.005;验证集OR=4.125,P=0.011)。在训练集和验证集中,综合诊断模型的预测能力的AUC值分别为0.879和0.849。结论影像组学模型的预测效能均高于常规模型,综合诊断模型明显优于常规模型,影像组学有望成为一种全新的生物学指标帮助临床预测NSCLC骨转移的风险。Objective To explore the value of the CT radiomics model in predicting bone metastasis in nonsmall cell lung cancer(NSCLC).Methods A retrospective analysis was performed on the clinical and imaging data of 149 patients diagnosed with NSCLC by percutaneous biopsy or surgical pathology in the hospital from December 2014 to December 2020.According to the examination results and clinical staging information,the patients were divided into a bone metastasis group(n=71)and a non-bone metastasis group(n=78).All patients were randomly divided into a training set(n=89)and a validation set(n=60)at a ratio of 6∶4.The ITK-SNAP software was used to manually delineate the region of interest(ROI)of the lesion layer by layer on the plain scan,arterial phase and venous phase images,and then 3-dimensional fusion was performed to obtain the volumetric ROI(VOI)of the lesion.The texture features of each lesion were then extracted by importing into the AK software.The minimum redundancy maximum relevance(mRMR)and the least absolute shrinkage and selection operator(LASSO)were used to screen the radiomics features of the plain scan,arterial phase,venous phase and multi-sequence(plain scan+arterial phase+venous phase)for the prediction of bone metastasis.A radiomics model was established using Logistic regression analysis,and the score(rad-score)of the each patient was calculated.The reliability was determined by the 100-time leave-P-out cross-validation method.Variables with statistically significant differences between groups in the clinical and imaging data were used to establish a conventional model through Logistic regression analysis.The conventional model was combined with the radiomics model with the highest predictive efficacy to establish a comprehensive diagnostic model,and a nomogram was drawn to analyze the predicted probability.The predictive ability of the model was evaluated using the receiver operating characteristic(ROC)curve,and the clinical application value of the model was evaluated using decision curve analysis(DC

关 键 词：非小细胞肺癌 骨转移 X线计算机技术 影像组学模型 

分 类 号：R445.3[医药卫生—影像医学与核医学]