基于影像学特征构建风险评估模型预测肺结节良恶性的研究

Establishment of a Risk Assessment Model Predicting Benign and Malignant Pulmonary Nodules Based on Imaging Features

目的:验证基于影像学特征的风险评估模型在胸外科肺结节手术患者中筛选肺恶性肿瘤的有效性。方法:采用回顾性研究,以2018年1月至2019年12月四川省肿瘤医院胸外科行手术治疗的351例患者为样本,包括278例肺恶性肿瘤患者和73例肺良性疾病患者。采集患者的影像学特征,并通过Logistic二分类回归分析,绘制受试者工作特征曲线,计算曲线下面积(area under curve,AUC),以Youden指数最大的分界点作为最佳诊断分界点。结果:结节直径每增加1 cm(OR=3.166,95%CI:1.983~5.055,P<0.001)、结节边缘有毛刺(OR=4.721,95%CI:2.487~8.962,P<0.001)、非实性结节(OR=6.392,95%CI:2.915~14.018,P<0.001)为肺恶性肿瘤的独立危险因素,差异具有统计学意义;基于影像学特征的风险评估模型Y=-1.618+1.153×X_(1)+1.552×X_(2)+1.855×X_(3),AUC为0.81(P<0.001),模型在Cut-off值为8.95时,对应的Youden指数为0.518,敏感度为0.669,特异度为0.849。结论:随着结节最大径的增加,恶性结节的概率增加;边缘有毛刺的结节较边缘无毛刺的结节恶性概率大;非实性结节较实性结节恶性概率大。基于影像学特征的风险评估模型可以提高术前判断肺结节良恶性的准确性。由此构建的模型具有一定的实用性,或可用于帮助临床医生判断肺结节的良恶性。

Objective:To verify the effectiveness of a risk assessment model based on imaging features in screening lung malignant tumors in patients undergoing thoracic surgery for pulmonary nodules.Methods:A retrospective study was conducted on 351 patients who underwent thoracic surgery in Sichuan Cancer Hospital from January 2018 to December 2019,including 278 patients with lung malignant tumors and 73 patients with benign lung diseases.We collected the imaging characteristics of patients,drew the receiver operating characteristic curve through binary logistic regression analysis,calculated area under the curve(AUC),and took the dividing point with the largest Youden index as the best cut-off point.Results:Every 1 cm increase in nodule diameter(OR:3.166,95%CI:1.983-5.055,P<0.001),burrs at the edge of the nodule(OR:4.721,95%CI:2.487-8.962,P<0.001)and non-solid nodules(OR:6.392,95%CI:2.915-14.018,P<0.001)were independent risk factors for lung malignancy,and the differences were statistically significant;AUC of the risk assessment model(Y=-1.618+1.153×X_(1)+1.552×X_(2)+1.855×X_(3))based on imaging features was 0.81(P<0.001).When the cut-off value was 8.95,the Youden index was 0.518,the sensitivity was 0.669,and the specificity was 0.849.Conclusion:The probability of malignant nodules increases as the maximum diameter of a nodule increases;nodules with burrs and non-solid nodules are more likely to be malignant.The risk assessment model based on imaging features can improve the accuracy of preoperative judgment of benign and malignant pulmonary nodules.