基于Gd-EOB-DTPA增强MRI的小肝癌微血管侵犯术前预测研究

Preoperative prediction of microvascular invasion in small hepatocellular carcinoma based on Gd-EOB-DTPA-enhanced MRI

目的探讨基于钆塞酸二钠(Gd-EOB-DTPA)增强MRI的临床—影像学特征模型对小肝癌微血管侵犯(MVI)的预测价值。方法将我院经术后病理诊断为小肝癌的174例患者按7∶3的比例随机分为训练组(n=122)和验证组(n=52)。由1名放射科医师分别在T2WI、门静脉期(PVP)、肝胆期(HBP)序列上进行病灶勾画,提取影像组学特征并对临床及影像学特征进行统计学分析;建立预测MVI的临床—影像学特征模型、影像组学特征模型及多模态(临床—影像学特征—影像组学特征)融合模型,计算3个模型的受试者工作特征(ROC)曲线下面积(AUC)、敏感度、特异度、准确度和F 1值,并评价模型对MVI的预测效能,采用Delong检验比较不同模型的ROC诊断结果。结果碱性磷酸酶(ALP)、凝血酶原时间(PT)、活化部分凝血活酶时间(APTT)、中性粒细胞、肿瘤假包膜、肿瘤形态、HBP瘤周低信号是预测小肝癌MVI的重要临床—影像学特征。PVP+HBP组合序列在基于影像序列的各模型中预测效能最佳(AUC=0.654)。临床—影像学特征模型预测MVI的效能最佳,AUC、准确度和F1值分别为0.748、0.846和0.868;临床—影像学特征模型的ROC诊断结果优于影像组学特征模型(P<0.05),临床—影像学特征模型与多模态融合模型的ROC诊断结果比较差异无统计学意义(P>0.05)。结论基于Gd-EOB-DTPA增强MRI建立的临床—影像学特征模型可较好地预测小肝癌MVI,但联合影像组学特征并不能提高其预测效能。

Objective To investigate the prediction value of the clinical-imaging feature model based on gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid(Gd-EOB-DTPA)-enhanced MRI for microvascular invasion(MVI)in small hepatocellular carcinoma.Methods A total of 174 patients with small hepatocellular carcinoma diagnosed by postoperative pathology in our hospital were randomly divided into the training group(n=122)and the verification group(n=52)according to the ratio of 7∶3.The lesions were delineated by one radiologist on T2WI,portal venous phase(PVP)and hepatobiliary phase(HBP)sequences,respectively,and the radiomics features were extracted,and the clinical and imaging features were analyzed statistically.The clinical-imaging feature model,the radiomics feature model and the multimodal(with clinical-imaging feature-radiomics feature)fusion model for the prediction of MVI were established.The area under the curve(AUC)of receiver operating characteristic(ROC),sensitivity,specificity,accuracy and F 1 value of the three models were calculated,and the prediction efficiencies of the models for MVI were evaluated.The ROC diagnostic results among different models were compared by Delong test.Results The alkaline phosphatase(ALP),prothrombin time(PT),activated partial thromboplastin time(APTT),neutrophils,tumor pseudocapsule,tumor shape and HBP peritumoral hypointensity were the important clinical-imaging features for the prediction of MVI in small hepatocellular carcinoma.The PVP and HBP combined sequences had the best prediction efficiency among the models based on imaging sequences(AUC=0.654).The clinical-imaging feature model had the best prediction efficiency for MVI,with the AUC,accuracy and F 1 value of 0.748,0.846 and 0.868,respectively;the ROC diagnostic result of the clinical-imaging feature model was better than that of the radiomics feature model(P<0.05).There was no significant difference in the ROC diagnostic results between the clinical-imaging feature model and the multimodal fusion model(P>0.05).Conclusion The clinical-imaging feature model established based on Gd-EOB-DTPA-enhanced MRI can preferably predict MVI in small hepatocellular carcinoma,while the prediction efficiency cannot be improved when combined with the radiomics features.