MRI放射组学模型预测肝细胞癌患者并发门静脉癌栓的价值  

The Value of MRI Radiomic Model in Predicting Portal Vein Tumor Thrombus in Patients with Hepatocellular Carcinoma

作　　者：牛稳 刘开才 邱晓晖[1] 刘艺超 吴兴旺[2] NIU Wen;LIU Kaicai;QIU Xiaohui(Medical Imaging Center,Bozhou Hospital of Anhui Medical University,Bozhou,Anhui Province 236800,P.R.China)

机构地区：[1]安徽医科大学附属亳州医院影像中心,236800 [2]安徽医科大学第一附属医院放射科,合肥230000

出　　处：《临床放射学杂志》2025年第5期888-892,共5页Journal of Clinical Radiology

基　　金：合肥市应用医学科研项目(编号:Hwk2022yb033)。

摘　　要：目的探讨MRI放射组学在预测肝细胞癌(HCC)并发门静脉癌栓(PVTT)的价值。方法回顾性选取经临床或病理证实的安徽医科大学第一附属医院和安徽医科大学附属亳州医院两个中心共258例HCC患者。根据MRI结果将患者分为PVTT阴性组168例和阳性组90例。中心1患者214例,按7∶3的比例随机分为训练组(n=150)和内部验证组(n=64);中心2患者44例,作为外部验证组。使用最小绝对收缩和选择算子(LASSO)算法进行降维和筛选特征,建立瘤体和瘤周放射组学模型。采用多因素Logistics回归分析筛选具有独立预测意义的影像学特征,再采用敏感度、特异度和受试者工作特征(ROC)曲线评估影像学模型、瘤体及瘤周放射组学模型和三者联合模型的效能。结果多因素回归分析显示肿瘤大小及强化包膜是PVTT发生的独立预测因子。联合模型于训练组、内部验证组及外部验证组的ROC曲线下面积(AUC)分别为0.876、0.839和0.841,均显著高于瘤周模型、瘤体模型及影像学特征模型,联合模型在训练组、验证组及外部验证组的敏感度(88.5%、82.0%和87.4%)及特异度(75.5%、81.8%和85.7%)均高于其他单一模型。结论基于MRI的瘤体及瘤周放射组学模型为早期预测HCC患者合并PVTT提供了有效的手段。Objective To explore the value of MRI radiomics in predicting portal vein tumor thrombus(PVTT)in patients with hepatocellular carcinoma(HCC).Methods A total of 258 HCC patients confirmed clinically or pathologically were retrospectively selected from two centers(the first affiliated hospital of Anhui medical university and bozhou hospital of Anhui medical university).Patients were divided into PVTT negative group(n=168)and positive group(n=90)according to MRI results.From center 1,214 patients were randomly divided into a training group(n=150)and an internal validation group(n=64)at a ratio of 7∶3.The 44 patients from center 2 were used as the external validation group.The least absolute shrinkage and selection operator(LASSO)algorithm was used to reduce dimensionality and screen features,and tumor and peritumoral radiomic models were established.Multivariate Logistic regression analysis was used to screen imaging features with independent predictive significance.Sensitivity,specificity,and receiver operating characteristic(ROC)curves were used to evaluate the efficacy of the imaging model,tumor radiomic model,peritumoral radiomic model,and the combined model.Results Multivariate regression analysis showed that tumor size and enhanced capsule were independent predictors of PVTT occurrence.The area under the curve(AUC)of the combined model in the training group,internal validation group,and external validation group were 0.876,0.839,and 0.841,respectively,which were significantly higher than those of the peritumoral model,tumor model,and imaging feature model.The sensitivity(88.5%,82.0%,and 87.4%)and specificity(75.5%,81.8%,and 85.7%)of the combined model in the training group,internal validation group,and external validation group were higher than those of other single models.Conclusion The tumor and peritumoral radiomic models based on MRI features provide an effective means for early prediction of PVTT in HCC patients.

关 键 词：放射组学 肝细胞癌 门静脉癌栓 

分 类 号：R735.7[医药卫生—肿瘤]