基于MR T2WI的影像组学对直肠癌新辅助治疗疗效的评估

Value of MR T2WI-based radiomics for the evaluation of the treatment response in rectal cancer after neoadjuvant therapy

目的:探讨基于高分辨T2WI的影像组学模型对评估直肠癌新辅助治疗疗效的价值。方法:回顾性分析2018年1月-2018年12月经手术病理证实且在接受新辅助治疗前、后均行MRI检查的80例直肠癌患者的病例资料。根据术后病理检查确定的肿瘤退缩分级(TRG),将TRG为0、1级者纳入疗效良好组,2、3级者纳入疗效不良组。在高分辨T2WI上勾画病灶的三维容积兴趣区(VOI)并使用两种模型提取影像组学特征,模型1:仅提取治疗前基线影像组学特征;模型B2:提取基线和治疗后的影像组学特征。随机选取70%的病例作为训练集,30%的病例作为测试集进行验证。对两种模型分别利用LASSO(least absolute shrinkage and selection operator)算法进行特征降维后,与TRG标签建立随机森林(RF)分类器,并分别进行受试者操作特征(ROC)曲线分析,比较两种模型的曲线下面积(AUC)并分析其诊断效能(敏感度、特异度、准确度、阳性预测值、阴性预测值、阳性似然比、阴性似然比)。采用决策曲线分析(DCA)评估临床获益。结果:模型1经降维后得到28个组学特征,模型2共获得3个组学特征,分别建立RF分类器模型,ROC曲线分析得到测试集模型1、2的AUC分别为0.943和0.950,两者间的差异无统计学意义(P>0.05)。模型1的特异度及阳性似然比较高,模型B的敏感度及阴性似然比较高。DCA显示总体上两种方法均可以临床获益。结论:基于治疗前及综合治疗前、后MR T2WI高分辨率图像的影像组学模型均可较准确地预测直肠癌新辅助治疗后的肿瘤退缩程度,可应用于临床上对直肠癌新辅助治疗疗效的评估。

Objective:The purpose of this study was to investigate the value of high-resolution T2WI-based radiomics for evaluating the treatment response in rectal cancer after neoadjuvant chemoradiotherapy(nCRT).Methods:The clinical and MRI data of 80 patients with pathologically confirmed rectal cancer who underwent MRI examination before and after nCRT from January 2018 to December 2018 in our hospital were retrospectively analyzed.According to postoperative tumor regression grade(TRG),the patients with TRG 0 and TRG 1 were included in the good response group;TRG 2 and TRG 3 were included in the poor response group.The volume of interest(VOIs)were drawn based on high-resolution T2WI images and then radiomics features were extracted from the VOIs by two models.Model 1:features were only extracted from baseline images before nCRT;Model 2:features were extracted from both images before and after nCRT.70%of the cases were randomly assigned to the training set and 30%to the validation set.The optimal features of the two models were selected by least absolute shrinkage and selection operator(LASSO)algorithm,and random forests(RF)classifiers was constructed with TRG signature and the receiver operating characteristic(ROC)curves were analyzed respectively.The area under the curve(AUC)and diagnostic efficacy indexes including sensitivity,specificity,accuracy,positive predictive value(PPV),negative predictive value(NPV),positive likelihood ratio(PLR)and negative likelihood ratio(NLR)of the two models were calculated and compared respectively.The decision curve analysis(DCA)was used to estimate clinic benefits.Results:28 optimal features of Model 1 and 3 optimal features of Model 2 were obtained.ROC analysis showed the AUC of Model 1 and 2 was 0.943 and 0.950 in the validation set,with no significant difference(P>0.05).Model 1 had higher specificity and PLR,meanwhile Model 2 had higher sensitivity and NPV.The DCA showed that both models could be beneficial for clinical practice in general.Conclusion:Radiomics based on MR high resolution T2WI images from baseline or baseline combined with nCRT can accurately predict the degree of tumor regression after neoadjuvant treatment for rectal cancer,thus can be used to evaluate the treatment response after nCRT in patients with rectal cancer.