Abstract:
The preoperative assessment of visceral pleural invasion (VPI) in patients with early lung adenocarcinoma is vital for surgical treatment. This study aims to develop and validate a CT-based radiomics nomogram to predict VPI in peripheral T1-sized solid lung adenocarcinoma. A total of 203 patients were selected as subjects, and were divided into a training cohort (n=141; scanned with Brilliance iCT256, Brilliance 64, Somatom Force, and Optima CT660) and a test cohort (n=62; scanned with Somatom Definition AS+). Radiomics characteristics were extracted from CT images. Variance thresholding, SelectKBest, and least absolute shrinkage and selection operator (LASSO) method were applied to determine optimum characteristics to construct the radiomic signature (radscore). After multivariate logistic regression analysis, a nomogram was structured regarding clinical factors, conventional CT features, and radscore. The nomogram property was tested based on its area under the curve (AUC). The nomogram based on the radscore and two conventional CT features (tumor pleura relationship and lymph node enlargement) showed high discrimination with an AUC of 0.877 (95% CI: 0.820-0.935) and 0.837 (95% CI: 0.737-0.937) in the training and test cohorts, respectively. The calibration curve and decision curve analysis showed good consistency and high clinical value of the nomogram. In conclusion, The CT-based radiomics nomogram was helpful in predicting VPI in peripheral T1-sized solid lung adenocarcinoma.
摘要:
术前评估早期肺腺癌患者的内脏胸膜侵犯(VPI)对于手术治疗至关重要。这项研究旨在开发和验证基于CT的放射组学列线图,以预测周围T1大小的实性肺腺癌的VPI。共选取203例患者作为研究对象,并分为一个训练组(n=141;用华晨iCT256、华晨64、SomatomForce扫描,和OptimaCT660)和一个测试队列(n=62;用Somatom定义AS+扫描)。从CT图像中提取影像组学特征。方差阈值,SelectKBest,应用最小绝对收缩和选择算子(LASSO)方法来确定构建放射学标记(radscore)的最佳特征。经过多因素logistic回归分析,列线图是关于临床因素的结构,常规CT特征,还有Radscore.基于其曲线下面积(AUC)测试列线图性质。基于radscore和两个常规CT特征(肿瘤胸膜关系和淋巴结肿大)的列线图显示出高度区分性,AUC为0.877(95%CI:0.820-0.935)和0.837(95%CI:0.737-0.937)在训练和测试队列中,分别。校准曲线和决策曲线分析显示,列线图具有良好的一致性和较高的临床价值。总之,基于CT的影像组学列线图有助于预测周围型T1大小的实性肺腺癌的VPI。