临床、CT特征及影像组学联合模型评估最大径≤2cm原发性肺腺癌侵袭性

Clinical,CT features and radiomics combined model for evaluating invasiveness of maximum diameter≤2 cm primary lung adenocarcinoma

目的观察临床、CT特征及影像组学联合模型评估最大径≤2 cm原发性肺腺癌侵袭性的价值。方法回顾性纳入116例最大径≤2 cm肺腺癌患者,依据病理结果将其分为高侵袭组(n=51)及低侵袭组(n=65),并按7∶3比例随机分为训练集(n=81)和测试集(n=35)。比较组间临床及CT特征差异;提取并筛选CT影像组学特征,计算影像组学评分(Rad-score)。采用logistic回归分析建立临床、CT特征及Rad-score联合模型,评估其预测效能及临床获益。结果组间患者恶性肿瘤家族史,病灶分叶征、毛刺征、存在提示预后不良病理表现占比差异均有统计学意义(P均<0.05);基于训练集数据筛选出8个影像组学特征并构建的联合模型评估训练集和验证集最大径≤2 cm肺腺癌侵袭性的曲线下面积分别为0.96和0.87,其预测值与真实值的误差较小,准确度较高,可使临床获益。结论临床、CT特征及影像组学联合模型可用于评估最大径≤2 cm原发性肺腺癌的侵袭性。

Objective To observe the value of a clinical,CT features and radiomics combined model for evaluating the invasiveness of maximum diameter≤2 cm primary lung adenocarcinoma.Methods Data of 116 patients with the maximum diameter≤2 cm lung adenocarcinoma were retrospectively analyzed.The patients were divided into highly invasive group(n=51)or low invasive group(n=65)based on postsurgical pathological findings,also randomly divided into the training set(n=81)or the test set(n=35)at the ratio of 7∶3.Clinical and CT features were compared between groups.Radiomics features were extracted and selected,and the radiomics scores(Rad-score)were calculated.Logistic regression analysis of the clinical,CT features and Rad-scores were performed to build a combined model,and the predictive performance and clinical benefits of this model were evaluated.Results Significant differences of family history of malignant tumors,as well as ratios of lobulation sign,spiculation sign and presence of pathological manifestations indicating poor prognosis were found between groups(all P<0.05).Based on training set,8 radiomics features were finally selected,and a combined model was constructed.The area under the curve for evaluating the invasiveness of maximum diameter≤2 cm lung adenocarcinoma was 0.96 in training set and 0.87 in test set.The error between predicted value and actual values of the above model was small,indicating high accuracy and potential clinical benefits.Conclusion The clinical,CT features and radiomics combined model could be used to evaluate the invasiveness of maximum diameter≤2 cm primary lung adenocarcinoma.