肺薄层CT特征预测纯磨玻璃结节浸润性腺癌的价值

The value of pulmonary thin-section CT characteristics in predicting invasive adenocarcinoma manifesting as pure ground-glass nodules

目的分析肺肿瘤性纯磨玻璃结节(pGGN)的薄层CT特征,构建逻辑(Logistic)回归模型预测浸润性腺癌的价值。方法回顾性收集2017年1月至2021年2月因胸部CT检查发现pGGN的连续病例,入组病例具有完整的临床影像资料和明确病理结果,依据病理类型分为腺体前驱病变组(不典型瘤样增生37例,原位腺癌51例)88例和浸润性病变组(微浸润腺癌84例,浸润性腺癌24例)108例。比较2组薄层CT特征分析指标是否存在差异,将具有统计学差异(P<0.05)的分析指标经二元多因素Logistic回归分析确定独立危险因素(P<0.05)并构建预测模型,绘制该模型及独立危险因素的受试者工作特征曲线(ROC),通过最大Youde指数确定最佳鉴别诊断效能。结果2组平均直径、平均CT值、肿瘤微血管CT成像征、空泡征、分叶征、毛刺征差异有统计学意义(P<0.05),2组空气支气管征差异无统计学意义(P=0.96)。Logistic回归预测模型表明肿瘤微血管CT成像征、分叶征是2组病理分类的独立预测因素(P<0.001),ROC曲线显示肿瘤微血管CT成像征、分叶征及Logistic回归预测模型的诊断效能均较高(AUC=0.88、0.8、0.74),并预测模型效能高于单项独立预测因素。结论肿瘤微血管CT成像征及分叶征是预测pGGN浸润性腺癌的重要CT特征,基于薄层CT特征Logistic回归模型有更高的鉴别诊断效能。

Objective To investigate thin-section computed tomography(CT) characteristics of pure ground-glass nodules(pGGNs),aiming to set up the logistic regression model in predicting the value of invasive adenocarcinoma(IA). Methods The consecutive pGGN cases underwent chest CT examinations were retrospectively enrolled, the complete clinical imaging data and clear pathological results of these patients were collcted, and the patients were assigned into the glandular precursor lesions group with 88 cases including 37 atypical neoplastic hyperplasia and 51 adenocarcinoma in situ, and invasive lesions group with 108 cases including 84 minimally invasive adenocarcinoma(MIA) and 24 IA. The differences in thin-section CT feature were compared between the two groups. The analysis indexes with statistical difference(P<0.05) by multivariate binary logistic regression analysis were confirmed to be independent risk factors(P<0.05) and set up a predictive model. The receiver operating characteristic curve(ROC) was drawn, and the top differential diagnosis performance was determined by the maximum Youde index.Results There were statistically significant in pathological classifications including average diameter, average CT value, tumor microvascular CT imaging sign, vacuole sign, lobular sign, and burr sign between two groups(P<0.05),while no statistical difference was found in the air bronchial sign(P=0.96). Logistic regression prediction model showed that tumor microvascular CT imaging sign and lobulation signs were independent predictors for differentiating the pathological classification(P<0.001). ROC curve showed that all diagnostic efficiencies of tumor microvascular CT imaging sign, lobulation sign and logistic regression prediction model were high(AUC=0.88,0.8,0.74),and the prediction model efficiency was higher than that of an independent predictor.Conclusion Tumor microvascular CT imaging sign and lobulation sign are important CT characteristics in predicting IA,manifesting as pGGNs, based on thin-section CT characteristics, Logistic regression model has higher differential diagnosis efficacy.