摘要
目的 回顾性分析结直肠管状绒毛状腺瘤(CRTVA)及CRTVA癌变患者的临床资料和增强CT图像,构建CRTVA癌变风险预测模型.方法 纳入病理诊断为CRTVA患者227例和CRTVA癌变患者125例,收集所有患者的增强CT图像并评估其特征.通过单因素分析和多因素Logistic回归分析筛选出相关临床和CT特征,分别构建临床模型、CT模型及临床-CT联合模型,绘制受试者工作特征(ROC)曲线并计算曲线下面积(AUC)进行模型评价.结果 352例患者的平均年龄为(63.56±11.55)岁(其中男180例).共收集和分析了 15个临床特征和12个CT影像特征.单因素分析显示其中3个临床特征[癌胚抗原(CEA)、粪便隐血试验(FOBT)、载脂蛋白A(Apo-A)]和10个影像特征(肿瘤部位、尺寸、血管、强化方式、边界、基底、形态、肠腔、肠壁、肿大淋巴结)在两组间有显著性差异(P＜0.05).多因素Logistic回归显示CEA、FOBT、血管、强化方式、边界、肠壁6个特征是CRTVA癌变的独立预测因素.构建3个预测模型,其中临床-CT联合模型表现出最佳预测性能,AUC 值为 0.951(95％CI:0.928～0.975;P＜0.001),高于临床模型的 AUC(0.683;95％CI:0.625～0.742;P＜0.001)和CT模型的AUC(0.942;95％CI:0.916～0.969;P＜0.001).结论 基于临床-CT特征构建的联合模型在预测CRTVA癌变风险方面表现出较高的诊断性能.
Abstract
Objective To construct a model for predicting the risk of malignant transformation of colorectal tubulovillous adenoma(CRTVA).Methods Contrast-enhanced CT of 227 patients with pathologically confirmed CRTVA and 125 patients with CRTVA carcinogenesis were retrospectively analyzed.Using univariate and multivariate logistic regression analysis of relevant clinical and CT features,clinical,CT,and clinical+CT models were constructed.The models were compared using area under the receiver operating characteristic curve(AUC)to calculate the sensitivity,specificity,accuracy,positive predictive value,and negative predictive value for predicting malignant transformation.Results The average age of 352 patients(180 men,172 women)was(63.56±11.55)years.A total of 15 clinical features and 12 CT features were analyzed.Univariate analysis showed significant differences between the two groups in 3 clinical features including carcinoembryonic antigen(CEA),fecal occult blood test(FOBT),apolipoprotein A(Apo-A),and 10 CT features including tumor location,size,blood vessels,contrast enhancement,boundary,base,morphology,intestinal lumen,intestinal wall,and enlarged lymph nodes(P＜0.05).Multivariate logistic regression analysis showed that 6 features including CEA,FOBT,blood vessels,contrast enhancement,boundary,and intestinal wall were independent predictive factors for CRTVA carcinogenesis,and three prediction models were constructed.The clinical+CT model showed the best predictive performance with an AUC value of 0.951(95％CI:0.928-0.975;P＜0.001)compared to the clinical model(0.683;95％CI:0.625-0.742;P＜0.001)and CT model(0.942;95％CI:0.916-0.969;P＜0.001).Conclusion The combined model constructed based on clinical and CT features has high performance in predicting the risk of CRTVA carcinogenesis.
维普
万方数据