比较超声造影（CEUS）和普美显磁共振成像（EOB-MRI）两者肝脏影像报告和数据系统（LI-RADS）对具有肝细胞癌（HCC）高危因素的≤3 cm的肝局灶性病变分类的一致性和诊断效能。

前瞻性收集天津市第三中心医院同时接受CEUS和EOB-MRI检查且具有HCC高危因素的结节直径≤3 cm的病例。在对患者信息、病理及相关检查均未知的情况下，对每个肝结节的CEUS及EOB-MRI图像进行观察，利用Kappa值评估观察者间一致性，分别按照CEUS LI-RADS（2017年版）及CT/MRI LI-RADS（2018年版）标准对结节进行分类。恶性病灶诊断以病理结果、良性病灶以病理结果和（或）随访作为参考标准，计算2种影像方式LI-RADS分类的诊断效能，利用χ²检验及Z检验比较两者诊断效能的差异。

107例114个结节满足入排标准。CEUS、EOB-MRI分类中LR-5、LR-4、LR-3和LR-M类结节的数量分别占45.6%、11.4%、19.3%、23.7%和25.4%、57.9%、10.5%、6.1%，2种影像方式对结节分类的一致性较差（Kappa=0.17，P<0.001）。CEUS LI-RADS和EOB-MRI LI-RADS分类各自观察者之间一致性良好（Kappa=0.83、0.77，P<0.001）。CEUS和EOB-MRI LR-5、LR-4、LR-3类对HCC的阳性预测值分别为98.1%、38.5%、13.6%和100%、71.2%、16.7%。CEUS和EOB-MRI LR-5类诊断HCC的敏感度、特异度、曲线下面积（AUC）分别为64.6%、97.1%、0.81和36.7%、100%、0.68。两者的阳性预测值、特异度差异均无统计学意义，AUC差异具有统计学意义（Z=4.29，P<0.01）。

CEUS和EOB-MRI对≤3 cm肝结节的分类一致性较差；两者各自LI-RADS分类不同观察者之间一致性较好。CEUS与EOB-MRI LR-5类对≤3 cm HCC具有同等良好的阳性预测值和特异度。

To compare the consistency and diagnostic performance of contrast-enhanced ultrasound (CEUS) versus gadoxetic acid-enhanced magnetic resonance imaging (EOB-MRI) in liver imaging reporting and data system (LI-RADS) classification of focal liver lesions ≤3 cm in patients at high risk of hepatocellular carcinoma (HCC).

Patients with focal liver lesions ≤3 cm at high risk of HCC who underwent CEUS and EOB-MRI at Tianjin Third Central Hospital were prospectively collected. The CEUS and EOB-MRI images of each liver nodule were observed to evaluate interobserver consistency using kappa coefficient and category according to the CEUS LI-RADS V2017 and CT/MRI LI-RADS V2018 criteria in a double blind manner. Pathology was used as the reference standard of malignant lesions and pathology and/or follow-up results were used as the reference standard for benign lesions to calculate the diagnostic performance of the two modalities, and χ² test and Z test were used to compare the difference between the two modalities.

A total of 114 nodules of 107 cases met the inclusion criteria. The proportions of LR-5, LR-4, LR-3, and LR-M diagnosed by CEUS and EOB-MRI accounted for 45.6%, 11.4%, 19.3%, and 23.7%, and 25.4%, 57.9%, 10.5%, and 6.1%, respectively. The consistency of nodule classification between the two modalities was poor (Kappa=0.17, P<0.001). There was good consistency between readers on CEUS LI-RADS and EOB-MRI LI-RADS (Kappa=0.83 and 0.77, P<0.001). The positive predictive values of CEUS and EOB-MRI LR-5, LR-4, and LR-3 for HCC were 98.1%, 38.5%, and 13.6%, and 100%, 71.2%, and 16.7%, respectively. The sensitivity, specificity, and area under the curve (AUC) of CEUS and EOB-MRI LR-5 for HCC were 64.6%, 97.1%, and 0.81, and 36.7%, 100%, and 0.68, respectively. The was no statistical difference in the positive predictive value or specificity between the two groups, while the difference of AUC was statistically significant (Z=4.29, P<0.01).

CEUS and EOB-MRI have poor consistency in classification of liver nodules ≤ 3 cm. The consistency between readers is good. CEUS and EOB-MRI LR-5 have equally good diagnostic performance for HCC ≤ 3 cm.