目的：基于MRI影像学特征探究肝脏特异性对比剂(Gd-EOB-DTPA)与常规对比剂(Gd-DTPA)在预测肝细胞癌(Hepatocellular carcinoma, HCC)微血管侵犯(Microvascular invasion, MVI)上的差异。方法：回顾性分析2010年1月至2019年1月行手术治疗的HCC患者253例(Gd-DTPA组121例、Gd-EOB-DTPA组132例)。评估患者术前MRI影像学特征,通过多因素logistics回归分别建立MVI预测模型,以受试者工作特征曲线及临床决策曲线评估模型效能及差异,并对患者术后生存情况进行随访分析。结果：Gd-DTPA组中病灶大小、边界侵袭征及Gd-EOB-DTPA组中病灶大小、边界侵袭征、卫星灶是预测MVI的独立风险因素。两组模型的曲线下面积(The area under the curve,AUC)分别为AUC(Gd-DTPA)：0.761、AUC(Gd-EOB-DTPA)：0.791。临床决策曲线显示Gd-DTPA与Gd-EOB-DTPA在预测MVI发生上具有相近的临床效益。生存分析显示组织病理证实的MVI阳性组与MVI阴性组的平均无疾病生存时间及模型预测的MVI阳性组与MVI阴性组的平均无疾病生存时间均存在显著差异。结论：病灶大小、病灶边界侵袭征及卫星灶是预测MVI的独立危险因素。使用Gd-EOB-DTPA在预测MVI发生方面不能比Gd-DTPA获得更多临床效益,但是Gd-EOB-DTPA具有较高的敏感性。
Objective: This study aims to explore the difference between hepatic-specific contrast agent (Gd-EOB-DTPA) and conventional contrast agent (Gd-DTPA) in predicting microvascular invasion (MVI) of hepatocellular carcinoma (HCC) based on MR imaging features. Methods: A retrospective analysis of 253 patients with HCC who underwent surgical treatment from January 2010 to January 2019 (121 cases in Gd-DTPA group and 132 cases in Gd-EOB-DTPA group) were carried out. The preoperative MR imaging features were evaluated, and MVI-prediction models were established through multi-factor logistics regression. Effectiveness and differences of the models were assessed with receiver operating characteristic curves and clinical decision curves. A postoperative survival analysis was also conducted. Results: Lesion size, border invasion in both Gd-DTPA and Gd-EOB-DTPA group, and satellite lesions in Gd-EOB-DTPA group, are independent risk factors for predicting MVI. The area under the curve (AUC) of two models were AUC(GD-DTPA): 0.761 and AUC(GD-EOB-DTPA): 0.791, respectively. The clinical decision curve shows that Gd-DTPA and Gd-EOB-DTPA have similar clinical benefits in predicting MVI. Survival analysis illustrated that there were significant difference on the average disease-free survival time between the MVI-positive group and the MVI-negative group not only confirmed by histopathology but also predicted by the models. Conclusion: Lesion size, border invasion and satellite lesions are independent risk factors for predicting MVI. The application of Gd-EOB-DTPA could not obtain more clinical benefits than Gd-DTPA in predicting MVI, but Gd-EOB-DTPA has a higher sensitivity.