基于MRI的放射组学模型预测中晚期肝癌患者TACE后肿瘤反应  

Application of MRI-based radiomics model in predicting the tumor response in patients with advanced hepatocellular carcinoma after transcatheter arterial chemoembolization

作　　者：顾萃华 潘仕文[2] 邱珏 王进进 GU Cuihua;PAN Shiwen;QIU Jue;WANG Jinjin(Department of Medical Imaging,Affiliated Suzhou Ninth Hospital of Soochow University(Suzhou Municipal Ninth People’s Hospital),Suzhou,Jiangsu Province 215200,China)

机构地区：[1]苏州大学附属苏州九院(苏州市第九人民医院)影像科,江苏苏州215200 [2]苏州大学附属第二医院影像科

出　　处：《介入放射学杂志》2024年第1期63-68,共6页Journal of Interventional Radiology

基　　金：江苏省第十五批“六大人才高峰”高层次人才选拔培养资助项目(WSW-223)。

摘　　要：目的基于临床特征和多模态磁共振成像(MRI)的放射组学纹理分析开发一款列线图模型,指导评估中晚期肝细胞癌(HCC)经导管动脉化疗栓塞(TACE)后3个月的肿瘤反应。方法纳入2017年1月至2021年12月苏州市第九人民医院病理确诊中晚期HCC患者105例,随机分为训练组63例和验证组42例。化疗前使用3.0T MRI采集T1WI、T2WI、动态对比剂增强(DCE)和弥散加权成像(DWI),A.K.软件提取纹理特征。化疗后3个月根据改良的实体瘤反应评估标准(m RECIST)分为反应组63例和无反应组42例。结果无反应组比反应组Child-Pugh分级B和BCLC分期C增多,血清甲胎蛋白(AFP)水平升高(P<0.05)。A.K.软件共提取MRI纹理特征396个,LASSO回归筛选最优预测因子6个,计算放射组学评分(Rad-score),ROC计算Rad-score预测训练组与验证组TACE后肿瘤无反应的AUC分别为0.842和0.803。多因素logistic回归模型显示,AFP≥50 ng/mL(OR=1.568,95%CI:1.234~1.902,P=0.003)、Child-Pugh分级B(OR=1.754,95%CI:1.326~2.021,P=0.001)、BCLC分期C(OR=1.847,95%CI:1.412~2.232,P=0.001)和Rad-score(OR=2.023,95%CI:1.569~2.457,P<0.01)是肿瘤无反应的独立危险因素。临床-放射组学结合预测肿瘤无反应的AUC最高。校正曲线显示列线图模型的吻合度较好。结论多模态MRI的放射组学纹理分析量化评分对预测中晚期HCC患者TACE后3个月的肿瘤无反应有一定价值,结合临床因素构建的列线图模型有较好的实践潜能。Objective To develop a nomogram model based on the clinical features and the radiomics texture analysis of multimodal magnetic resonance imaging(MRI),so as to predict the tumor response in patients with advanced hepatocellular carcinoma(HCC)3 months after receiving transcatheter arterial chemoembolization(TACE).Methods A total of 105 patients with advanced HCC,whose diagnosis was pathologically-confirmed at the Suzhou Municipal Ninth People’s Hospital between January 2017 and December 2021,were enrolled in this study.The patients were randomly divided into training group(n=63)and verification group(n=42).Before chemotherapy,T1WI,T2WI,dynamic contrast-enhanced(DCE)scan,and diffusion-weighted imaging(DWI)were performed by using a 3.0T MRI scanner.A.K.software was used to extract the texture.Three months after chemotherapy,according to the modified Response Evaluation Criteria in Solid Tumors(mRECIST)the patients were divided into response group(n=63)and non-response group(n=42).Results Compared with the response group,in the non-response group the percentage of Child-Pugh grade B and BCLC stage C was obviously higher and the serum alpha fetoprotein(AFP)level was remarkably elevated(P<0.05).A.K.software extracted 396 MRI texture features,and LASSO regression analysis screened out 6 optimal predictors.The radiation score(Rad-score)was calculated by ROC.The AUC of Rad-score for predicting tumor non-response after TACE by ROC in the training group and verification group were 0.842 and 0.803 respectively.Multivariate logistic regression model analysis showed that AFP≥50 ng/mL(OR=1.568,95%CI=1.234-1.902,P=0.003),Child-Pugh grade B(OR=1.754,95%CI=1.326-2.021,P=0.001),BCLC stage C(OR=1.847,95%CI=1.412-2.232,P=0.001)and Rad-score(OR=2.023,95%CI=1.569-2.457,P<0.001)were the independent risk factors for tumor non-response.Clinico-radiomics combination had the highest AUC value for predicting tumor nonresponse.The correction curve showed that the nomogram model had a good agreement.Conclusion The quantitative score of radi

关 键 词：磁共振成像 放射组学 纹理分析 列线图 肝细胞癌 经导管动脉化疗栓塞 肿瘤反应 

分 类 号：R735.7[医药卫生—肿瘤]