基于非增强CT影像组学列线图在术前预测胸腺上皮肿瘤风险分类中的应用价值  

Application value of radiomics nomogram based on non-enhanced CT for the preoperative prediction of risk classification of thymic epithelial tumor

作　　者：张梦琪[1] 玉苏甫·肉孜 洪悦[1] 陈杰[1] ZHANG Mengqi;Yusufu Rouzi;HONG Yue;CHEN Jie(Department of Radiology and Medical Imaging,People’s Hospital of Xinjiang Uygur Autonomous Region,Urumqi 830001,China)

机构地区：[1]新疆维吾尔自治区人民医院放射影像中心,新疆乌鲁木齐830001

出　　处：《中国中西医结合影像学杂志》2023年第6期687-692,714,共7页Chinese Imaging Journal of Integrated Traditional and Western Medicine

摘　　要：目的:基于非增强CT(NECT)影像组学列线图在术前预测胸腺上皮肿瘤(TET)风险分类中的应用价值。方法:回顾性收集92例确诊为TET患者的临床和影像资料,根据TET的WHO简化病理分型,分为低危组(A、AB、B1型)48例、高危组(B2、B3、C型)44例;按照7∶3的比例,随机分成训练集(64例)和测试集(28例)。勾画病灶ROI并进行影像组学特征的提取,降维和筛选后,选择最优特征,构建9种机器学习模型并选择最优模型,对TET风险进行预测;采用单因素分析的方法筛选临床特征建立临床模型。结合临床评分、影像组学评分构建联合模型,以列线图进行可视化;使用ROC曲线、DeLong检验及校准曲线评估模型性能。结果:共提取影像组学特征1834个,降维、筛选后获得11个组学特征。选取支持向量机(SVM)构建影像组学模型。选取纵隔肿大淋巴结、胸膜/心包肥厚建立临床模型。临床模型、影像组学模型及联合模型在测试集中的AUC值分别为0.812、0.812、0.875,绘制列线图,校准曲线显示联合模型具有较好的一致性。结论:在预测TET病理分型中,基于所选临床特征及NECT影像组学特征的可视化列线图模型可能具有良好的临床应用前景。Objective:To explore the application value of radiomics nomogram based on non-enhanced CT(NECT)for the preoperative prediction of risk classification of thymic epithelial tumor(TET).Methods:A total of 92 patients with TET confirmed by pathology were retrospectively enrolled.According to the WHO simplified pathological classification of TET,48 patients were divided into a low-risk group(A,AB,B1 type),and 44 patients were divided into a high-risk group(B2,B3,C type).92 cases were randomly divided into the training cohort(64 cases)and the validation cohort(28 cases)at a ratio of 7∶3.After delineating the ROI and extracting radiomic features,followed by dimension reduction and selection,the optimal features were chosen.9 machine learning models were constructed to predict the risk of TET.Single factor analysis was used to screen clinical features to establish a clinic model.Based on clinic score and radiomics score,the combined model was established and visualized by nomogram.The performance of the three models was evaluated using ROC curve,DeLong test and calibration curve.Results:A total of 1834 radiomics features were obtained,and 11 features were selected to construct the radiomics model with SVM.The clinic model was established with enlarged mediastinal lymph nodes and pleural or pericardial hypertrophy.The AUC of the clinic model,radiomics model and the combined model in the validation cohort were 0.812,0.812 and 0.875,respectively.A nomogram model based on clinic score and radiomics score was established,and the calibration curve showed the combined model had a good consistency.Conclusion:The visual nomogram model based on the selected clinical features and NECT radiomics features may have good clinical application prospects in predicting the WHO simplified pathological classification of TET.

关 键 词：胸腺上皮肿瘤 WHO病理分型 影像组学 支持向量机 列线图 体层摄影术 X线计算机 

分 类 号：R736.3[医药卫生—肿瘤] R730.44[医药卫生—临床医学]