临床特征联合CT影像组学列线图术前预测直肠癌脉管侵犯的价值

The value of clinical features combined with radiomics nomogram analysis based on enhanced CT venous phase in preoperative prediction of lympho-vascular invasion of rectal cancer

目的:探讨临床特征联合增强CT静脉期影像组学构建的列线图术前预测直肠癌脉管侵犯(LVI)的价值。方法:选取皖南医学院弋矶山医院2017年6月至2022年6月经组织病理学确诊的299例直肠癌患者为训练组,164例患者为验证组。根据LVI状态分为LVI阳性组与LVI阴性组。使用MaZda软件基于增强CT静脉期图像提取纹理特征,使用最小绝对收缩和选择算子(LASSO)回归对纹理特征进行降维,并建立影像组学标签,计算标签评分(Rad-score)。采用单因素及多因素Logistic回归分析筛选独立危险因素,建立预测LVI状态的模型,并绘制模型的列线图。使用受试者工作特征(ROC)曲线及曲线下面积(AUC)评价模型的区分度,使用Hosmer-Lemeshow检验及校准曲线评价模型的拟合优度及校准度,使用决策曲线(DCA)评价模型的临床适用性。结果:多因素Logistic回归分析发现癌胚抗原(CEA)、肿瘤分化程度及Radscore为独立危险因素,列线图在训练集AUC为0.811 (95%CI:0.760~0.861),验证集AUC为0.761 (95%CI:0.676~0.845)。Hosmer-Lemeshow检验结果,训练组(χ^(2)=7.981,P=0.435),验证组(χ^(2)=4.050,P=0.853),均无统计学意义。校准曲线显示列线图预测结果与实际情况一致性好。DCA表明该模型具有较高的临床适用性。结论:临床特征联合CT影像组学列线图术前预测直肠癌LVI状态具有较高的效能。

Objective:To investigate the value of clinical features combined with radiomics nomogram analysis based on enhanced CT venous phase in preoperative prediction of lympho-vascular invasion(LVI)of rectal cancer.Methods:A total of 299 patients with rectal cancer diagnosed by histopathology in Yijishan Hospital of Wannan Medical College from Jun 2017 to Jun 2022 were enrolled as the training group,and 164 patients were enrolled as the verification group.They were divided into LVI positive group and LVI negative group according to LVI status.MaZda software was used to extract texture features based on enhanced CT venous phase images,and least absolute shrinkage and selection operator(LASSO)regression was used to reduce the dimension of texture features,and radiomic signature was established to calculate the Rad-score.The independent risk factors were screened by univariable and multivariable logistic regression analysis.And the prediction model of LVI status was established,and the nomogram of the model was drawn.Receiver operating characteristic(ROC)curve and area under curve(AUC)were used to evaluate the model's discrimination.Hosmer-Lemeshow test and calibration curve were used to evaluate the model's goodness of fit and calibration,and decision curve(DCA)was used to evaluate the model's clinical applicability.Results:Multivariable logistic regression analysis showed that carcinoembryonic antigen(CEA),tumor differentiation status and Rad-score were independent risk factors.The AUC of the nomogram was 0.811(95%CI:0.760-0.861)in the training group and 0.761(95%CI:0.676-0.845)in the verification group.Hosmer-Lemeshow test obtained an insignificant result in the training group(χ2=7.981,P=0.435)and the verification group(χ2=4.050,P=0.853)The calibration curve showed that the predicted results of nomogram were in good agreement with the actual situation. DCAshowed that this model had high clinical applicability. Conclusion:The combination of clinical features and radiomics nomogramanalysis based on enhanced CT has high efficacy in preoperative prediction of LVI status of rectal cancer.