术前T2磁共振影像组学在预测介入治疗大肝癌近期疗效的研究  

Preoperative T2 MRI radiomics signature in predicting the short-term efficacy of interventional therapy for large HCC: a clinical study

作　　者：孙跃军 白洪林 王栋 冉昭 徐李刚 钱晟 杨国威 张巍 王建华 高欣[2] 刘嵘 SUN Yuejun;BAI Honglin;WANG Dong;RAN Zhao;XU Ligang;QIAN Sheng;YANG Guowei;ZHANG Wei;WANG Jianhua;GAO Xin;LIU Rong(Department of Interventional Radiology,Affiliated Zhongshan Hospital,Fudan University,Shanghai 200032,China)

机构地区：[1]复旦大学附属中山医院介入科,上海200032 [2]中国科学院苏州生物医学工程技术研究所医学影像部 [3]上海市影像医学研究所

出　　处：《介入放射学杂志》2019年第11期1036-1041,共6页Journal of Interventional Radiology

摘　　要：目的探讨基于介入术前MR的影像组学对肝癌6个月内介入治疗局部反应的预测价值。方法回顾性分析70例接受介入治疗的大肝癌患者,术前1周内行MRI检查,术后1~2个月、半年随访多功能MRI,按照mRECIST标准对介入术后6个月肿瘤局部控制率进行评估,利用MITK软件在T2图像上进行勾画肿瘤区域,提取影像组学特征,利用Pearson相关系数剔除冗余特征,利用mRMR特征排序方法筛选得到组学标签。采用留一法划分样本,与LASSO分类器共同构建大肝癌介入治疗后6个月进展预测模型。计算模型的受试者工作特征曲线(ROC)的曲线下面积(AUC),评估模型的预测能力。结果 70例患者6个月后随访52例病灶未进展(N-PD),18例进展(PD),两组术前T2-MRI的Wavelet-HHH firstorder Mean等3个的影像组学参数有统计学差异(P<0.05),以这些差异性影像组学特征所建的模型具有一定的预测能力(AUC=0.657,ACC=0.714,SEN=0.588,SPE=0.755)。结论基于MRI组学特征能对接受介入治疗的大肝癌进行近期进展风险预测,可在TACE术前筛选进展危险性高的患者,为其提早采取联合治疗提供个体化建议。Objective To evaluate preoperative T2 MRI radiomics signature in predicting the local response of large hepatocellular carcinoma(HCC) to transcatheter arterial chemoembolization(TACE) within 6 months. Methods The clinical data of 70 patients with large HCC who receive interventional therapy were retrospectively analyzed. One week before TACE, MRI was performed in all patients. Follow-up examination with multi-functional MRI was carried out within 1-2 months and at 6 months after TACE. According to mRECIST criterion, the 6-month local control rate of tumor was evaluated. By using MITK software, the tumor areas were delineated on T2 images, and the radiomics signature features were extracted. Pearson correlation coefficient was used to eliminate redundant features. The mRMR feature ranking method was used to screen the radiomics features, and radiomics label was thus obtained. The samples were divided by leaving one method, which, together with LASSO classifier, was used to construct the prediction model for 6-month post-TACE progress of large HCC. The receiver operating characteristic curve(ROC) and the area under curve(AUC) of the computational model were calculated, and the results were used to evaluate the prediction ability of the model. Results The 70 patients were followed up for 6 months, non-progression disease(NPD) was seen in 52 patients and progression disease(PD) was seen in 18 patients. Statistically significant differences in three preoperative radiomics parameters, including Wavelet-HHH first order mean, etc. on T2-MRI, existed between the two groups(P<0.05). The model that was constructed with the above mentioned differential radiomics characteristics had certain prediction ability(AUC =0.657, ACC =0.714, SEN =0.588,SPE=0.755). Conclusion Based on the characteristics of MRI radiomics, it is possible to predict the shortterm progress risk for large HCC treated with TACE. This technique can be used to screen patients with high risk of progression before TACE, and to provide individualized advice for

关 键 词：大肝癌 介入治疗 疗效预测 T2磁共振影像组学 

分 类 号：R735.7[医药卫生—肿瘤]