BACKGROUND We aimed to develop radiomic models based on different phases of computed tomography (CT) imaging and to investigate the efficacy of models for diagnosing mediastinal metastatic lymph nodes (LNs) in non-small cell lung cancer (NSCLC). METHODS Eighty-six NSCLC patients were enrolled in this study, and we selected 231 mediastinal LNs confirmed by pathology results as the subjects which were divided into training (n = 163) and validation cohorts (n = 68). The regions of interest (ROIs) were delineated on CT scans in the plain phase, arterial phase and venous phase, respectively. Radiomic features were extracted from the CT images in each phase. A least absolute shrinkage and selection operator (LASSO) algorithm was used to select features, and multivariate logistic regression analysis was used to build models. We constructed six models (orders 1-6) based on the radiomic features of the single- and dual-phase CT images. The performance of the radiomic model was evaluated by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV). RESULTS A total of 846 features were extracted from each ROI, and 10, 9, 5, 2, 2, and 9 features were chosen to develop models 1-6, respectively. All of the models showed excellent discrimination, with AUCs greater than 0.8. The plain CT radiomic model, model 1, yielded the highest AUC, specificity, accuracy and PPV, which were 0.926 and 0.925; 0.860 and 0.769; 0.871 and 0.882; and 0.906 and 0.870 in the training and validation sets, respectively. When the plain and venous phase CT radiomic features were combined with the arterial phase CT images, the sensitivity increased from 0.879 and 0.919 to 0.949 and 0979 and the NPV increased from 0.821 and 0.789 to 0.878 and 0.900 in the training group, respectively. CONCLUSIONS All of the CT radiomic models based on different phases all showed high accuracy and precision for the diagnosis of LN metastasis (LNM) in NSCLC patients. When combined with arterial phase CT, the sensitivity and NPV of the model was be further improved.

中文翻译：

---

基于不同时相CT影像学特征对NSCLC纵隔转移淋巴结的鉴别诊断


背景我们的目的是开发基于计算机断层扫描（CT）成像不同阶段的放射组学模型，并研究模型诊断非小细胞肺癌（NSCLC）纵隔转移淋巴结（LN）的功效。方法 86例NSCLC患者纳入本研究，选取病理结果证实的231例纵隔淋巴结作为研究对象，分为训练组（n = 163）和验证组（n = 68）。分别在平扫期、动脉期和静脉期的 CT 扫描上勾画出感兴趣区域 (ROI)。从每个阶段的 CT 图像中提取放射组学特征。使用最小绝对收缩和选择算子（LASSO）算法来选择特征，并使用多元逻辑回归分析来构建模型。我们根据单相和双相 CT 图像的放射组学特征构建了六个模型（1-6 阶）。放射组学模型的性能通过受试者工作特征曲线下面积（AUC）、敏感性、特异性、准确性、阳性​​预测值（PPV）和阴性预测值（NPV）进行评估。结果每个ROI总共提取了846个特征，分别选择10、9、5、2、2和9个特征来开发模型1-6。所有模型都表现出出色的辨别力，AUC 大于 0.8。平扫 CT 放射组学模型（模型 1）产生最高的 AUC、特异性、准确性和 PPV，分别为 0.926 和 0.925； 0.860 和 0.769； 0.871 和 0.882；训练集和验证集分别为 0.906 和 0.870。当平扫期和静脉期 CT 放射组学特征与动脉期 CT 图像相结合时，敏感性从 0.879 和 0.919 增加到 0。训练组中的 949 和 0979，NPV 分别从 0.821 和 0.789 增加到 0.878 和 0.900。结论 基于不同时相的CT影像组学模型对于NSCLC患者LN转移（LNM）的诊断均表现出较高的准确度和精密度。与动脉期CT结合，模型的敏感性和NPV进一步提高。