首页|钆塞酸二钠增强MRI对单发肝细胞癌微血管侵犯的预测价值

钆塞酸二钠增强MRI对单发肝细胞癌微血管侵犯的预测价值

Predictive Value of Gd-EOB-DTPA Enhanced MRI in Microvascular Invasion of Single Hepatocellular Carcinoma

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
目的 探讨临床资料、基于钆塞酸二钠(Gd-EOB-DTPA)增强MRI的影像征象及影像组学方法对单发肝细胞癌(HCC)患者微血管侵犯(MVI)的单独及联合预测价值.方法 110 例HCC患者(训练集:77 例,验证集:33 例),分析入组患者的临床资料,评估MRI影像征象,并分别从动脉期、门静脉期和延迟期 3 个序列中提取瘤内影像组学特征,使用最小绝对收缩和选择算子(LASSO)算法进行特征选择,随后使用Logistic回归分析构建临床模型、影像征象模型、单序列影像组学模型、影像组学联合模型和临床-影像征象-影像组学模型,并评估不同模型的性能.结果 110 例患者中,48 例(43.6%)经病理证实存在MVI.在训练集中,年龄在MVI状态方面差异有统计学意义(P<0.05).在训练集及验证集中,γ-谷氨酰转肽酶(γ-GT)、动脉期瘤周强化、肝胆期瘤周低信号都是MVI的有力预测因子(P<0.05).通过LASSO算法,最终选择了17 个与MVI相关的影像组学特征.在 3 个单序列影像组学模型中,基于门静脉期图像建立的影像组学模型对HCC患者MVI的预测能力表现最佳,其在训练集和验证集中的曲线下面积(AUC)分别为0.788、0.699.最终构建的临床-影像征象-影像组学模型在训练集中表现出较高的效能,相应AUC为0.895(95%CI:0.827～0.963),特异度为0.744,敏感度为0.912.结论 本研究建立了一种基于患者临床资料,Gd-EOB-DTPA增强MRI的影像征象及影像组学多个序列特征的联合模型,并验证了其在术前无创预测HCC患者MVI的价值,其有可能作为指导后续个体化治疗的临床工具.
Objective To explore the single and combined predictive value of clinical data,gadolinium-ethoxybenzyl-di-ethylenetriamine pentaacetic acid(Gd-EOB-DTPA)enhanced MRI features and radiomics methods for microvascular inva-sion(MVI)in patients with single hepatocellular carcinoma(HCC).Methods This study included 110 patients with HCC(training set:77 cases,validation set:33 cases).The clinical data of the patients were analyzed,the MRI imaging fea-tures were evaluated,and the intratumoral imaging features were extracted from arterial phase,portal venous phase and de-layed phase respectively.The least absolute shrinkage selection operator(LASSO)algorithm was used for feature selection.Then logistic regression analysis was used to construct clinical model,imaging model,single sequence radiomics model,com-bined radiomics model and clinical-imaging-radiomics model,and evaluate the performance of different models.Results Of the 110 patients,48 cases(43.6%)were confirmed by pathology to have MVI.In the training set,there were significant differences in age in MVI state(P<0.05).In the training set and validation set,γ-glutamyl transpeptidase(γ-GT)、peritu-moral enhancement on arterial phase,peritumoral hypointensity on hepatobiliary phase were strong predictors of MVI(P<0.05).Through the LASSO algorithm,this study finally selected 17 radiomics features related to MVI.Among the three sin-gle sequence radiomics models,the radiomics model based on portal vein phase images had the best ability to predict the MVI of HCC patients,and its AUC in training set and validation set is 0.788 and 0.699 respectively.Finally,the clinical-imaging-radiomics model showed high efficiency in the training set,with AUC of 0.895(95%CI:0.827～0.963),specifici-ty of 0.744 and sensitivity of 0.912.Conclusion This study established a combined model based on clinical data,Gd-EOB-DTPA enhanced MRI imaging features and multiple sequences radiomics features,and verified its value of non-invasive prediction of MVI in patients with HCC before operation,which may be used as a clinical tool to guide follow-up individual-ized treatment.

Hepatocellular carcinomaMicrovascular invasionMagnetic resonance imagingGadolinium-ethoxy-benzyl-diethylenetriamine pentaacetic acidRadiomics

王少怡、周智鹏、蒋宇、林斌、方行、李晨、王玲廷、刘蒋静

展开 >

541000 桂林医学院附属医院放射科

肝细胞癌 微血管侵犯 磁共振成像 钆塞酸二钠 影像组学

广西肝脏损伤与修复分子医学重点实验室项目桂林市科学研究与技术开发计划重大专项

GXLIRMMKL-K20201020190202-2

2024

临床放射学杂志
黄石市医学科技情报所

临床放射学杂志

CSTPCD北大核心
影响因子：0.872
ISSN：1001-9324
年,卷(期)：2024.43(4)
  • 18