CT影像组学对前纵隔小结节的诊断价值

The Diagnostic Value of CT Radiomics in Small Anterior Mediastinal Nodules

摘要: 目的：评估从胸部增强CT图像中提取的影像组学特征对胸腺上皮源性肿瘤和囊肿的鉴别能力，建立诺模图协助诊断前纵隔小结节的病理性质。方法：回顾性分析2019年1月至2022年12月在青岛大学附属医院因前纵隔小结节而接受手术治疗的患者，并分为胸腺上皮性源肿瘤与囊肿性病变两组。从术前胸部增强CT的动脉期与静脉期图像中分别提取病灶的影像组学特征进行筛选和降维后构建影像组学得分(Rad-score)，结合临床与影像形态学特征共同进行单因素与多因素分析。以多因素Logistic回归有意义的独立预测因子构建诊断模型及对应列线图。通过受试者工作特征曲(ROC)、校准曲线及决策曲线分析(DCA)对列线图的诊断能力进行评估。结果：共244例患者纳入本研究，其中TET 89例，囊肿性病变155例。多因素回归结果显示结节有无强化、结节形状及Rad-score是病理性质的独立预测因子。基于这些因素构建的列线图在训练集(AUC 0.95)和测试集(AUC 0.86)上均显示出较好的诊断效能，校准曲线和DCA显示了列线图较好的诊断准确性和临床应用价值。结论：通过影像组学特征构建并验证了一个前纵隔小结节诊断列线图，作为一种无创诊断工具可以比较准确地区分胸腺上皮性肿瘤和囊肿性病变。

Abstract: Objectives: The aim of this study is to assess the efficacy of radiomics features extracted from contrast-enhanced chest CT scans in distinguishing between TET (Thymic Epithelial Tumors) and cystic lesions, as well as establishing a nomogram to aid in diagnosing the pathological properties of SMAN (Small Anterior Mediastinal Nodules). Methods: A retrospective analysis was conducted on patients who underwent surgical treatment for SAMN at the Affiliated Hospital of Qingdao University from January 2019 to December 2022. The patients were divided into two groups: TET and cystic lesions. The radiomics features of the nodules were extracted from arterial phase and venous phase images obtained through enhanced chest CT scans prior to surgery. After feature screening and dimensionality reduction, a Rad-score was constructed. Univariable and multivariable analyses were performed in conjunction with clinical, radiologic characteristics and Rad-score. An independent predictor-based multifactor Logistic regression model was developed along with a corresponding nomogram for diagnosis. The diagnostic ability of the nomogram was assessed using receiver operating characteristic curves (ROC), calibration curves, and decision curve analysis (DCA). Results: A total of 244 patients were included in this study, including 89 TETs and 155 cystic lesions. The results of multivariable regression analysis demonstrated that enhancement, shape, and Rad-score were independent predictors of the pathological nature of nodules. The nomogram constructed based on these factors exhibited superior diagnostic performance in both the training set (AUC 0.95) and the test set (AUC 0.86). Furthermore, the calibration curve and decision curve analysis (DCA) revealed enhanced diagnostic accuracy and clinical applicability of the nomogram. Conclusion: Based on radiomics features, a diagnostic nomogram of SAMN was constructed and validated, which can be used as a noninvasive diagnostic tool to accurately distinguish TET from cystic lesions.