临床-影像组学列线图术前预测直肠癌T分期

Preoperative prediction of T-stage of rectal cancer by clinical features and radiomics nomogram

目的:探讨临床特征联合影像组学列线图术前预测直肠癌T分期的价值。方法:回顾性分析2018年1月至2022年7月收治的508例经组织病理学确诊的直肠癌患者的临床及影像资料。将患者按7∶3随机分为训练组(n=355)及验证组(n=153)。用MaZda软件基于增强CT静脉期图像分别提取瘤体及瘤周纹理特征,利用LASSO回归对训练组纹理特征降维,建立影像组学标签,并计算标签评分。采用多因素Logistic回归分析筛选独立危险因素,构建临床-影像组学联合预测模型,并绘制模型的列线图及校准曲线,使用受试者工作特征(ROC)曲线评价各模型的预测效能,决策曲线(DCA)评价模型的临床适用性。结果:基于瘤体及瘤周共提取614个纹理特征,降维后得到6个最优特征。建立的所有模型中列线图效能最高。列线图的AUC在训练组为0.894(95%CI:0.860~0.928),验证组为0.946(95%CI:0.914~0.978)。结论:临床-影像组学列线图术前能够有效地预测直肠癌T1-2期与T3-4期,并将预测结果可视化,为临床直肠癌规范化治疗提供参考依据。

Objective:To explore the value of clinical features combined with radiomics nomogram in predicting T-stage of rectal cancer before operation.Methods:Clinical data of 508 rectal cancer patients diagnosed with histopathology from Jan 2018 to Jul 2022 were retrospectively analyzed.The patients were randomly divided into training group(n=355)and validation group(n=153)at a ratio of 7∶3.MaZda software was used to extract the texture features of tumor and adjacent tissues based on enhanced CT venous phase images.LASSO regression was used to reduce the dimension of texture features of training group,the radiomics signature was established and the signature score(Rad-score)was calculated.Multivariable logistic regression analysis was used to screen independent risk factors.A clinical-radiomics combined prediction model was constructed,and the nomogram and calibration curve of the model was drawn.ROC curve was used to evaluate the prediction efficiency of each model,and decision curve analysis(DCA)was used to evaluate the clinical applicability of the model.Results:A total of 614 texture features were extracted based on the tumor and adjacent tissues,and 6 optimal features were obtained after dimension reduction.The nomogram was the most effective of all the established models.The AUC of nomogram in training group was 0.894(95%CI:0.860-0.928),and that in validation group was 0.946(95%CI:0.914-0.978).Conclusion:The clinical-radiomics nomogram can effectively predict the stage T1-2 and stage T3-4 of rectal cancer before operation,and visualize the prediction results,which provides reference for standardized treatment of rectal cancer.