增强MR组学联合临床因素预测肝细胞癌根治性切除术后早期复发的价值研究

The study of enhanced MR radiomics combining clinical factors in predicting early recurrence of hepatocellular carcinoma after resection

目的探究联合临床因素的肝细胞癌(hepatocellular carcinoma,HCC)术前MR影像组学模型对预测HCC切除术后早期复发的价值。材料与方法回顾性分析116例(训练集82例、测试集34例)术前进行过腹部MRI扫描且经病理确诊为HCC患者的动态对比增强MRI(dynamic contrast enhanced MRI,DCE-MRI)的图像和临床因素。运用3D Slicer软件勾画病变感兴趣区(region of interest,ROI)并提取影像组学特征,通过最大相关-最小冗余算法、最小绝对值收缩和选择算子(least absolute shrinkage and selection operator,LASSO)降维,使用LASSO建立影像组学评分,引入临床因素构建Logistic回归模型,并联合影像组学评分构建列线图模型,通过计算受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under the curve,AUC),进行Delong检验和临床决策曲线分析(decision curve analysis,DCA),评估和比较每个模型的预测性能和模型之间差异是否具有统计学意义。结果共选取9个影像组学特征构建影像组学模型。临床因素包括TNM分期(美国癌症联合委员会第8版癌症分期),甲胎蛋白水平,γ-谷氨酰氨基转移酶,Child-Pugh分级。联合影像组学评分和临床因素的影像组学列线图的鉴别效能优于临床因素模型[AUC=0.79(95%CI:0.63~0.96)vs.AUC=0.71(95%CI:0.52~0.90),DeLong检验Z=2.363,P=0.018],两个ROC曲线差异有统计学意义,经DCA验证其临床净收益率更高。结论基于术前MR影像组学及临床因素的联合模型对于预测HCC切除术后早期复发具有应用价值。

Objective:To develop and validate a preoperative MRI radiomics model combining clinical factors in predicting early recurrence of hepatocellular carcinoma after surgical resection.Materials and Methods:One hundred and sixteen patients(82 in the training set and 34 in the test set),who had been pathologically diagnosed as hepatocellular carcinoma(HCC)with preoperative abdominal dynamic contrast-enhancement magnetic resonance imaging(DCE-MRI)and relevant clinical factors,were recruited in this retrospective study.The 3D slicer software was used to delineate the ROI of lesions and extract the radiomics features.The radiomics score model was established by utilizing the maximum correlation-minimum redundancy algorithm(mRMR),minimum absolute contraction and selection operator(LASSO)feature selection procedure,Similarly,the clinical factors were introduced to build the Logistic regression model.The area under the receiver operating characteristic curve(AUC),Delong test and decision curve analysis(DCA)were performed to evaluate and compare the accuracy and difference of each radiomics model.Results:In all,nine radiomics features were selected to construct the radiomics score model.The clinical factors model,including TNM stage,alpha-fetoprotein level,γ-glutamylaminotransferase,Child-Pugh grade.The radiomics nomogram of integrated the radiomics score and clinical factors demonstrated better discriminative performance(AUC=0.79,95%CI:0.63-0.96)than the Clinical factors models(AUC=0.71,95%CI:0.52-0.90),Delong test Z=2.363,P=0.018.The decision curve analysis presented the improved clinical net benefit.Conclusion:The combined model based on preoperative MR radiomics and clinical factors can be served as effective imaging biomarker to predict early recurrence of hepatocellular carcinoma after surgical resection.