血管分型特征对磨玻璃结节型肺腺癌侵袭程度的预测价值

Value of vascular characteristics in predicting invasiveness of lung adenocarcinoma manifesting as ground glass nodule

目的探讨高分辨率CT(HRCT)上血管分型特征对磨玻璃结节(GGN)型肺腺癌侵袭程度的预测价值。方法回顾性分析我院2018年1月—2019年12月经手术病理证实为肺腺癌的649枚磨玻璃结节。依据术后病理类型将其分为非侵袭性腺癌组[原位腺癌(AIS)/微浸润性腺癌(MIA)](292例)与侵袭性腺癌组[浸润性腺癌(IAC)](357例)。分析GGN的HRCT征象以及GGN内部和周围的血管特征。采用二元Logistic回归分析建立预测模型,并通过受试者工作特征(ROC)曲线和Z检验比较各模型的诊断效果。结果两组之间具有显著统计学意义的HRCT征象包括支气管充气征、分叶征、毛刺征、空泡征、胸膜凹陷征、结节直径大小、肿瘤实性成分占比(CTR)、血管分型特征(P<0.001)。加入血管分型特征联合其他CT独立预测变量建模与未加入血管分型特征建立的预测模型相比,AUC值为0.853(P<0.001,95%CI:0.83~0.88)比0.814(P<0.001,95%CI:0.78~0.85),敏感度为72.3%比74.8%,特异度为82.2%比76.4%,提高了特异度和AUC值。Z检验ROC曲线下面积之差为0.039(P<0.05)。结论血管分型特征联合其他HRCT特征建立综合预测模型,有助于提高GGN型肺腺癌侵袭程度的预测效能。

Objective To investigate the value of blood vessel typing on high-resolution CT(HRCT)in predicting the invasion degree of ground glass nodule(GGN)lung adenocarcinoma.Methods A retrospective analysis was performed on 649 ground glass nodules pathologically confirmed as lung adenocarcinoma in our hospital between January 2018 and December 2019.According to postoperative pathological types,the patients were categorized into non-invasive adenocarcinoma group[(adenocarcinoma in situ(AIS)/minimally invasive adenocarcinoma(MIA)](n=292)and invasive adenocarcinoma group[invasive adenocarcinoma(IAC)](n=357).HRCT signs of GGN and vascular features within and surrounding GGN were analyzed.Binary logistic regression analysis was used to establish the prediction model,and the diagnostic performance of each model was compared using receiver operating characteristic(ROC)curve and Z test.Results The HRCT findings that showed statistical significance between the two groups included air bronchogram,lobular sign,burr sign,vacuole sign,pleural sag sign,diameter size,consolidation-to-tumor ratio(CTR),and vascular classification characteristics(P<0.001).When comparing the combined prediction model composed of all other independent CT predictors with vascular classification to the model without vascular classification,the AUC value of the former was 0.853(P<0.001,95%CI:0.83-0.88)compared to 0.814(P<0.001,95%CI:0.78-0.85)for the latter;with respective sensitivities of 72.3%and 74.8%,as well as specificities of 82.2%and 76.4%.These results indicate an improvement in specificity and AUC value by incorporating vascular classification into the model analysis(P<0.05).Conclusion The combination of vascular classification features with other HRCT features enables the establishment of a comprehensive prediction model,which can be valuable in predicting the degree of invasion efficiency in GGN lung adenocarcinoma.