基于术前MRI动脉期的影像组学对肝细胞肝癌射频消融术后短期疗效的预测分析  

Prediction analysis of short-term therapeutic efficacy after radiofrequency ablation for hepatocellular carcinoma based on preoperative arterial phase MRI radiomics

作　　者：伍伦鑫 刘迎春 王欧成 尧麒 张海熠 汪静 刘勇 WU Lunxin;LIU Yingchun;WANG Oucheng;YAO Qi;ZHANG Haiyi;WANG Jing;LIU Yong(School of Clinical Medicine,Southwest Medical University,Luzhou,Sichuan Province 646000,China;MR Room,the Affiliated Traditional Chinese Medicine Hospital of Southwest Medical University,Luzhou,Sichuan Province 646000,China;Department of Radiology,Hejiang People’s Hospital,Hejiang,Sichuan Province 646200,China;Department of Hepatobiliary Diseases,the Affiliated Traditional Chinese Medicine Hospital of Southwest Medical University,Luzhou,Sichuan Province 646000,China)

机构地区：[1]西南医科大学临床医学院,四川泸州646000 [2]西南医科大学附属中医医院MR室,四川泸州646000 [3]合江县人民医院放射科,四川合江646200 [4]西南医科大学附属中医医院肝胆病科,四川泸州646000

出　　处：《实用放射学杂志》2024年第8期1281-1285,共5页Journal of Practical Radiology

基　　金：四川省重点研发计划项目(2019YFS0018);国家中医临床研究基地建设单位科研项目(西南医大中医院【2020】33号)。

摘　　要：目的探讨基于术前MRI动脉期的纹理分析构建肝细胞肝癌(HCC)射频消融(RFA)术后短期疗效预测模型的可行性。方法回顾性分析169例经RFA治疗的HCC患者。根据术后短期疗效将患者分为预后好组(112例)及预后差组(57例)。利用Mazda软件提取患者术前MRI动脉期图像的纹理特征,通过Fisher系数、交互信息、分类误差概率和平均相关系数降维。按7︰3的比例将患者分为训练组(n=119)和测试组(n=50),利用独立样本t检验+最小绝对收缩和选择算子(LASSO)算法进一步进行特征筛选,最后采用LASSO回归建立影像组学模型,并通过受试者工作特征(ROC)曲线及曲线下面积(AUC)评估模型的预测效能。结果影像组学模型的组成特征为S_2__2_SumOfSqs、Teta1、S_5_0_DifVarnc、S_2_0_DifEntrp、Horzl_LngREmph、S_5_5_InvDfMom。模型在训练组、测试组的AUC分别为0.987[95%置信区间(CI)0.965~1.000]、0.918(95%CI 0.818~1.000),敏感度分别为98.7%(95%CI 92.4~100)、93.9%(95%CI 84.8~100),特异度分别为97.5%(95%CI 90.0~100)和88.2%(95%CI 70.6~100)。结论基于术前MRI动脉期的纹理分析构建HCC经RFA术后短期疗效的预测模型是可行的,且具有良好的预测效能。Objective To explore the feasibility of constructing a short-term therapeutic efficacy prediction model for hepatocellular carcinoma(HCC)after radiofrequency ablation(RFA)based on texture analysis of preoperative MRI arterial phase images.Methods A retrospective analysis was conducted on 169 HCC patients treated with RFA.Based on the short-term therapeutic efficacy,the patients were divided into a good prognosis group(112 cases)and a poor prognosis group(57 cases).Texture features of preoperative MRI arterial phase images were extracted using Mazda software,and dimension reduction was performed through Fisher coefficient,mutual information,classification error probability,and mean correlation coefficient.The patients were divided into a training group(n=119)and a testing group(n=50)in a 7︰3 ratio.Independent sample t-tests and the least absolute shrinkage and selection operator(LASSO)algorithm were employed for further feature selection.Subsequently,a radiomics model was established using LASSO regression and evaluated through the receiver operating characteristic(ROC)curve and area under the curve(AUC).Results The radiomics model comprised features such as S_2__2_SumOfSqs,Teta1,S_5_0_DifVarnc,S_2_0_DifEntrp,Horzl_LngREmph,and S_5_5_InvDfMom.The AUC of the model were 0.987[95%confidence interval(CI)0.965-1.000]and 0.918(95%CI 0.818-1.000)in the training and testing groups,respectively.The sensitivity was 98.7%(95%CI 92.4-100)and 93.9%(95%CI 84.8-100),and the specificity was 97.5%(95%CI 90.0-100)and 88.2%(95%CI 70.6-100),respectively.Conclusion The construction of a predictive model for short-term therapeutic efficacy of HCC after RFA based on texture analysis of preoperative MRI arterial phase images is feasible and demonstrates good predictive performance.

关 键 词：磁共振成像 影像组学 肝细胞肝癌 射频消融 

分 类 号：R445.2[医药卫生—影像医学与核医学] R445[医药卫生—诊断学] R735.7[医药卫生—临床医学]