分享:
分享到微信朋友圈
X
综述
钆塞酸二钠增强MRI预测肝细胞癌微血管侵犯的研究进展
陈智英 谭艳

Cite this article as CHEN Z Y, TAN Y. Research progress of gadoxetic acid disodium-enhanced MRI in predicting microvascular invasion of hepatocellular carcinoma[J]. Chin J Magn Reson Imaging, 2026, 17(8): 206-211, 225.本文引用格式 陈智英, 谭艳. 钆塞酸二钠增强MRI预测肝细胞癌微血管侵犯的研究进展[J]. 磁共振成像, 2026, 17(8): 206-211, 225. DOI:10.12015/issn.1674-8034.2026.08.024.


[摘要] 肝细胞癌(hepatocellular carcinoma, HCC)是世界上常见的肿瘤之一,其治疗后复发率高,预后差。微血管侵犯(microvascular invasion, MVI)作为HCC侵袭性生物学行为的核心标志,是影响HCC患者术后复发和生存的重要因素。目前,MVI诊断主要依靠术后病理活检,因有创及滞后,无法在术前为临床医生提供患者MVI风险等级的精准评估,进而无法指导术前新辅助治疗方案的制定,同时也不能为手术方式的选择及术中手术切缘的设定提供指导,无助于术前规划与术中决策。传统影像学检查(如CT、常规超声)对MVI的显示敏感度较低,钆塞酸二钠(gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid, GD-EOB-DTPA)作为肝细胞特异性对比剂,其影像征象可有效评估MVI风险,对于术前无创预测HCC MVI具有重要作用。在此基础上发展的人工智能技术为HCC术前MVI预测提供了新路径。本文突破既往综述“泛化覆盖多类影像学技术”的局限,以GD-EOB-DTPA增强MRI为核心主线,从GD-EOB-DTPA增强MRI影像征象及相关影像组学、生境成像和深度学习预测HCC MVI的研究现状展开综述,为HCC患者提供个性化治疗,为临床诊断和治疗提供参考。
[Abstract] Hepatocellular carcinoma (HCC) is one of the most common tumors worldwide, with a high recurrence rate and poor prognosis after treatment. Microvascular invasion (MVI), as a core marker of the aggressive biological behavior of HCC, is an important factor affecting the postoperative recurrence and survival of HCC patients. Currently, the diagnosis of MVI mainly relies on postoperative pathological biopsy, which is invasive and lagging, and cannot provide a precise assessment of the MVI risk level for patients before surgery for clinicians, thus failing to guide the formulation of preoperative neoadjuvant treatment plans. It also cannot provide guidance for the selection of surgical methods and the setting of surgical margins during the operation, and is of no help for preoperative planning and intraoperative decision-making. Traditional imaging examinations (such as CT and conventional ultrasound) have a low sensitivity in showing MVI. Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (GD-EOB-DTPA), as a liver cell-specific contrast agent, its imaging signs can effectively evaluate the MVI risk and play an important role in non-invasive prediction of HCC MVI before surgery. On this basis, the development of artificial intelligence technology provides a new path for the preoperative prediction of MVI in HCC. This article breaks through the limitations of previous reviews that "generalize and cover multiple imaging techniques", and takes GD-EOB-DTPA-enhanced MRI as the main thread. This article will review the current research status of GD-EOB-DTPA enhanced MRI imaging signs and related radiomics, habitat imaging and deep learning in predicting HCC MVI, providing personalized treatment for HCC patients and references for clinical diagnosis and treatment.
[关键词] 肝细胞癌;微血管侵犯;钆塞酸二钠;影像组学;深度学习;生境分析
[Keywords] hepatocellular carcinoma;microvascular invasion;GD-EOB-DTPA;radiomics;deep learning;habitat analysis

陈智英 1   谭艳 2*  

1 山西医科大学医学影像学院,太原 030001

2 山西医科大学第一医院影像科,太原 030001

通信作者:谭艳,E-mail: tanyan123456@sina.com

作者贡献声明::谭艳拟定本综述的写作思路,指导撰写稿件,并对稿件重要内容进行了修改,获得了国家自然科学基金项目资金支持;陈智英起草和撰写稿件,获取、分析并解释本综述的参考文献;全体作者都同意最后的修改稿发表,都同意对本研究的所有方面负责,确保本综述的准确性和诚信。


基金项目: 国家自然科学基金项目 82371941
收稿日期:2026-03-30
接受日期:2026-07-11
中图分类号:R445.2  R735.7 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2026.08.024
本文引用格式 陈智英, 谭艳. 钆塞酸二钠增强MRI预测肝细胞癌微血管侵犯的研究进展[J]. 磁共振成像, 2026, 17(8): 206-211, 225. DOI:10.12015/issn.1674-8034.2026.08.024.

0 引言

       肝细胞癌(hepatocellular carcinoma, HCC)是世界上常见的肿瘤之一,其治疗后复发率高,预后差[1]。微血管侵犯(microvascular invasion, MVI)是显微镜下可见的血管内皮细胞衬覆的脉管腔内癌细胞团[2],作为HCC侵袭性生物学行为的核心标志,是影响HCC患者术后复发和生存的重要因素。其术前精准评估对临床治疗方案制定、个体化干预及预后判断具有重要指导价值。目前,MVI主要依靠术后病理诊断,具有滞后性,传统影像学检查(如CT、常规超声)对MVI的显示敏感度较低,难以满足术前无创精准预测的临床需求。钆塞酸二钠(gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid, GD-EOB-DTPA)作为一种肝细胞特异性MRI对比剂,可通过动态增强扫描获取动脉期(arterial phase, AP)、门静脉期(portal venous phase, PVP)、平衡期及肝胆特异期(hepatobiliary phase, HBP)多期相影像信息,清晰呈现肿瘤形态、血供特点,可有效评估MVI风险,对于术前无创预测HCC MVI具有重要作用[3, 4]。在此基础上发展的人工智能(artificial intelligence, AI)技术为HCC术前MVI预测提供了新路径[5, 6, 7]。本文突破既往综述“泛化覆盖多类影像学技术”的局限,以GD-EOB-DTPA增强MRI为核心主线,从GD-EOB-DTPA增强MRI影像征象及相关影像组学、生境成像和深度学习(deep learning, DL)预测HCC MVI的研究现状展开综述,以期为临床实现无创预测手段提供参考,帮助临床医生选择合适的治疗方案,改善患者预后,提高患者生存率。

1 文献检索

       为系统阐述GD-EOB-DTPA增强MRI在HCC术前MVI预测中的研究现状,本文检索国内外数据库,包括中国知网、万方数据库、维普和PubMed,检索时间为2019年至2026年3月,该阶段相关影像诊断、模型分析技术趋于成熟,研究成果贴合本课题研究方向,同时统一时间边界,确保文献分析具备可比性,使用主题词及自由词检索,中文检索式为(“钆塞酸二钠”OR“普美显”OR“GD-EOB-DTPA”)AND(“肝细胞癌”OR“肝癌”OR“HCC”)AND(“微血管侵犯”OR“MVI”);英文检索式为(“gadoxetate disodium”OR“GD-EOB-DTPA”)AND(“Hepatocellular Carcinoma”OR“HCC”)AND(“Microvascular Invasion”OR“MVI”),共检索到文献455篇,经过去重、文章题目和摘要判断、全文阅读后,排除419篇文献。排除标准:(1)研究对象接受其他治疗,如抗肿瘤化疗、靶向免疫治疗等;(2)不可获取完整数据的文献;(3)样本量<20例的研究;(4)重复发表的文献,个案报告、评述、综述或会议摘要类文献等。最终纳入文献36篇。

2 GD-EOB-DTPA增强MRI影像征象预测HCC的MVI

       相较于传统MRI及其他影像学检查[8],GD-EOB-DTPA增强MRI凭借其肝细胞特异性对比剂的独特优势,可更清晰地显示HCC的病灶特征。超声造影(contrast-enhanced ultrasound, CEUS)虽具有成本低、易获得、可实时监测等优势,但HUANG等[8]对比研究表明,GD-EOB-DTPA增强MRI预测HCC的MVI中特异性更高,达到90.5%。

2.1 MRI影像征象对MVI的预测价值

       HONG等[9]通过荟萃分析36项相关研究,筛选出7项与MVI显著相关的MRI影像特征:肿瘤体积较大(>5 cm)、AP边缘强化、AP瘤周强化、HBP瘤周低信号、肿瘤边缘不规则、多灶性病变以及T1加权成像(T1WI)低信号。HBP瘤周低信号和多发病灶对MVI诊断的特异性分别达91.1%和93.3%。LU等[10]研究亦证实,HBP瘤周低信号是MVI的独立预测因子(OR=4.343,P=0.003)。这一结果可能与MVI发生时,HBP肿瘤周围肝细胞的有机阴离子转运多肽(OATPs)表达下降,导致对比剂摄取减少密切相关[11]。此外,BEI等[12]构建的模型的受试者工作特征曲线下面积(area under the curve, AUC)为0.871,提示,AP边缘强化、肿瘤包膜不完整和肿瘤内血管增强是MVI的独立风险因子。上述模型效能良好,但一致性有待验证,且多基于回顾性研究,临床转化需结合临床数据调整。

2.2 LI-RADS评分体系辅助优化MVI预测效能

       针对人工阅片征象判读标准不统一的问题,临床将标准化LI-RADS评分体系与GD-EOB-DTPA增强MRI影像征象结合,有效提升MVI预测效能。YANG等[13]表明,结合肝脏影像报告和数据系统2018版(Liver Imaging Reporting And Data System, LI-RADS v2018)标准,晕环状强化、马赛克征、肿瘤边缘不光滑及HBP瘤周低信号这四项影像学特征,可作为预测MVI的影像学标志物,任意一项阳性时,MVI敏感度可达90%以上。改良版LI-RADS(mLI-RADS)在维持高特异性的同时显著提升小HCC诊断敏感性[14],更高效地通过影像征象(包括扩散加权成像扩散受限、HBP低信号、马赛克结构)捕捉MVI相关影像学线索,进一步优化MVI预测效能。

2.3 HBP成像时间对MVI预测模型效能的影响

       GD-EOB-DTPA增强MRI的HBP成像时间是影响病灶特征显示的关键扫描参数,同时影响预测模型的整体诊断效能。ZHANG等[15]基于HBP的5 min、10 min和15 min采集的图像构建的预测模型AUC分别为0.685、0.718和0.795,HBP 15 min可作为模型构建的最优延迟时间。然而,最优延迟时间的选择并非绝对,需大样本、多中心研究进一步细化选择标准,为临床诊疗决策提供可靠支撑。

       值得注意的是,传统MRI阅片主要依赖视觉提取肿瘤影像特征,主观性较强,在MVI预测中存在局限性;有研究发现,人工阅片对MVI诊断的AUC均<0.75[16],表明人工阅片的评估方式主观性较强,存在一定争议,为弥补单一影像特征预测的不足、规避人工阅片的主观偏差,ZHANG等[17]将影像特征与临床信息联合建立模型,预测MVI的特异性达到93.2%。结果证实融合影像特征及临床客观数据构建的预测模型,亦能提高术前MVI预测效能。

3 影像组学预测HCC的MVI

       尽管上述影像征象预测模型取得了良好效能,但人工阅片主观性强、一致性差(AUC<0.75)[16],且难以量化肿瘤内部细微异质性。影像组学[18, 19]通过高通量特征提取和机器学习建模,为MVI预测提供了更客观、定量的解决方案。

       目前,基于GD-EOB-DTPA增强MRI术前预测HCC的MVI的影像组学研究已有较多报道。FENG等[20]仅基于HBP图像提取定量特征构建MVI术前预测模型(AUC=0.83)。在此基础上,WANG等[21]系统研究7个MRI序列,结果表明脂肪抑制的T1加权图像(AUC=0.877)和HBP(AUC=0.862)的预测效果显著优于其他序列。PENG等[22]进一步证实,多期相联合模型(AUC=0.889)显著高于单相(AUC=0.789)、两相(AUC=0.815)及三相(AUC=0.848)模型的预测性能。此外,ZENG等[23]基于AP、PVP和HBP的iTED特征开发MiTED-AP、MiTED-PVP和MiTED-HBP三种影像组学模型,并通过提取AP与PVP特征差值获取时间特征构建增量模型(MDelta),研究证实Delta影像组学时间特征预测MVI表现出超越传统AP、PVP的特征。以上研究分别从单期相、多序列、多期相联合及时间维度特征等层面展开探索,为该领域预测模型的构建奠定了重要基础。

       目前,预测HCC的MVI的模型种类繁多,但单一影像模型存在局限性,仅依赖图像数据,难以全面反映HCC的生物学特征。WANG等[24]构建基于多模态MRI和临床数据的模型(AUC=0.968),展现出良好的预测效能。与之类似,WANG等[25]构建整合炎症信息、图像特征及多机器学习的融合模型(AUC范围:0.809~0.908),预测能力显著优于单一影像组学模型。多项研究[26, 27, 28]亦证实,影像组学联合临床影像学特征的综合模型,在MVI预测中具有可靠性。综上,基于GD-EOB-DTPA增强MRI影像组学模型对术前预测MVI具有一定的价值,且融合临床特征的模型性能更优。但需重视模型可解释性不足与标准化缺失的问题,在可解释性方面,多数AI预测模型存在“黑箱”困境,仅能输出预测结果,无法明确模型决策核心依据的影像特征及生物学意义。梯度加权类激活映射(gradient-weighted class activation mapping, Grad-CAM)通过可视化模型关注影像区域[29],明确与MVI相关的关键病灶部位(如肿瘤边缘、强化不均匀区域),直观呈现模型决策的空间依据;SHAP(SHapley Additive exPlanations)量化各影像特征对预测结果的贡献度[30, 31],筛选对MVI预测有价值的核心特征(高阶纹理特征、基于小波的特征)。在模型中,Grad-CAM负责“视觉定位”,SHAP负责“特征量化”,二者从空间视觉定位与特征贡献量化两个维度,全面提升MVI预测模型的临床可解释性与结果可信度。在标准化方面,影像组学特征提取缺乏统一的行业标准,软件平台、图像重建参数(如层厚、视野、矩阵大小)及感兴趣区(region of interest, ROI)勾画方法的差异,导致特征提取不一致、研究结果可比性差及模型泛化能力不足。因此,建立统一的影像组学标准化流程是其走向临床应用的重要条件。

       此外,由于影像特征筛选以及模型的构建方法差异,模型的预测性能可能也不尽相同。CHEN等[32]证实HBP为MVI预测的最优时相,其构建的支持向量机(support vector machine, SVM)、极端梯度提升(eXtreme Gradient Boosting, XGBoost)及逻辑回归(logistic regression, LR)模型均表现出极高的预测效能(AUC分别为0.942、0.938、0.936),三者均被推荐为潜在的有效模型。此外,郭剑波等[33]采用主成分分析法对特征进行降维并构建SVM预测MVI模型(AUC=0.92)。目前,学术界尚未明确MVI预测的最佳模型,现有模型存在样本偏倚、泛化能力不足、外部验证匮乏等短板,需大样本量、多中心研究进一步探索。总之,基于GD-EOB-DTPA增强MRI影像组学特征、结合多种算法构建的模型具有重要应用价值;融合临床及多模态、多序列MRI信息的模型,可进一步提升预测性能。

4 生境成像预测HCC的MVI

       传统影像组学通常将肿瘤看作一个整体来提取高通量定量特征,难以体现肿瘤内部的异质性。生境分析(habitat analysis, HA)将肿瘤划分为多个具有不同功能或结构特征的亚区[34],在技术上通过生境亚区聚类对各区域独立建模,弥补影像组学难以捕捉局部微环境差异的不足。且近年来多项研究[35, 36]指出,瘤周区域可为反映肿瘤的异质性提供潜在重要信息。

       肿瘤本身具有异质性,由多个具备不同生物学特征的亚区组成,被称为肿瘤微环境,而肿瘤内异质性(intratumoral heterogeneity, ITH)的量化分析可有效提升术前预测MVI的准确性。ZHANG等[37]基于多种MRI序列,根据信号强度、纹理特征和空间结构的相似性,采用生境聚类+变异系数量化构建模型,多中心数据验证其预测效能优异(AUC=0.82~0.99)。同样,蒋韬[38]提取灰度共生矩阵、灰度游程矩阵、灰度区域矩阵等纹理特征及局部熵聚类量化肿瘤异质性,结果表明基于肿瘤边缘亚区模型(AUC=0.884)预测性能优于肿瘤内部区域(AUC=0.769)。综上,ITH量化分析在MVI预测中具有重要价值。

       鉴于约85%的MVI发生于肿瘤边缘1 cm以内,相关研究重点分析了肿瘤周围组织衍生的影像学特征[39]对MVI预测的价值。FENG等[20]提取瘤内和瘤周区域影像定量特征,发现瘤内联合瘤周的模型(AUC=0.85)预测效能显著优于单一瘤内或瘤周模型,这也是首次基于GD-EOB-DTPA增强MRI建立瘤内联合瘤周MRI影像组学模型预测HCC的MVI的研究。同样,ZHANG等[40]构建用于MVI预测的列线图模型,5折交叉验证的AUC范围从0.880到1.000,显示出高诊断准确性。YU等[41]的研究也显示联合模型(AUC=0.925)显示高可靠性。现阶段部分联合预测模型虽表明瘤内异质性联合瘤周1 cm区域特征的预测优势[20, 42],但不同模型效能仍存在显著差异,主要归因于样本量大小、病例纳入排除标准、模型构建算法等研究设计差异。从研究类型而言,目前,该领域研究以回顾性研究为主[43],前瞻性研究相对匮乏。CHEN等[44]采用黄金角径向稀疏并行技术开展前瞻性研究,定量评价与MVI相关的HCC瘤内和瘤周血管动力学改变,构建的综合模型(AUC=0.817)虽性能略低于部分回顾性研究,但也为模型的临床转化提供了有力支撑。

       总体上,基于GD-EOB-DTPA增强MRI瘤内联合瘤周模型术前预测HCC的MVI展现出良好的性能。部分研究[45]进一步纳入三级淋巴结构,其通常与预后和更强的免疫治疗反应相关,作用机制可能是促进淋巴细胞浸润、肿瘤抗原激活及分化,以增强抗肿瘤免疫反应[46]。李依蔓等[47]通过构建“影像结构异质性+免疫微环境异质性”的联合预测模型,模型预测性能较好(AUC=0.825),有望为HCC术后辅助治疗奠定基础,为患者提供个性化诊疗。目前,多数研究存在样本量偏小、外部验证不足等短板,模型泛化能力与临床落地实用性有待考量。后续研究需大样本量,去进一步验证联合模型的有效性和稳健性,为MRI数据深度挖掘以及MVI的预测提供更多可能。

5 DL预测HCC的MVI

       相较于传统影像组学依赖人工手动提取特征的局限性,DL技术通过构建卷积神经网络(convolutional neural networks, CNN)模型,实现影像特征的自动提取,在影像分割[48]、信息分类[19]和预测研究[49]等方面的应用中展现出卓越的性能。已有研究[50, 51]证实其在HCC术前MVI预测中的效能优于传统影像组学。

       作为MVI预测基础的DL架构,传统CNN依托卷积核完成影像局部特征提取,但不同空间维度的CNN模型在临床应用中利弊分化明显,需结合临床数据特点合理决策。3D模型在捕捉复杂3D解剖结构方面优势显著,而2D模型则具备更高的计算效率[52],二者各有优劣。WANG等[53]研究表明2D-expansion-DL模型(AUC=0.70)与3D-DL模型(AUC=0.72)预测效能相近,且2D-expansion-DL模型更易实现,提示模型的选择应贴合实际临床数据需求。另有研究[54]指出,2.5D模型既保留2D模型轻量化、运算高效、适配低算力设备的优势,又有效弥补单一层面特征缺失问题,完整捕获病灶层间连续性特征,降低过拟合风险。LUO等[55]基于HBP图像特征构建的模型显示,2.5D DL模型预测效能最佳,验证集和测试集的AUC值为0.802和0.759,其性能与3D DL模型相当甚至更优,进一步佐证了2.5D DL模型的临床适用性,与WANG等[53]的研究结论形成互补。

       尽管上述模型在MVI预测中取得可观效果,但受限于卷积操作的局部感受野,对全局信息的捕捉能力存在局限,难以充分挖掘图像中与MVI相关的全域特征。相比之下,基于Transformer的架构[56]通过自注意力机制突破该局限,更优量化肿瘤全域异质性。HE等[57]构建Transformer的多模态融合模型(AUC=0.886),验证了模型的有效性。黄倩等[58]研究表明,Swin Transformer模型凭借分层自注意力与全局建模能力,在MVI预测中均显著优于传统CNN模型及Vision Transformer(ViT)模型。特别是在延迟期(delayed phase, DP)AUC高达0.993,且结合Grad-CAM技术可直观验证其关注区域与病理MVI高度吻合,表明模型预测结果的可靠性。CNN与Transformer混合架构模型兼具局部特征提取与全局依赖建模的双重优势,克服单一架构的缺陷,提升预测性能。

       现阶段大多研究围绕MVI单一病理指标开展独立预测,而临床中MVI和肿瘤簇包埋血管(vessels that encapsulate tumor clusters pattern, VETC)均为HCC的高危病理特征,二者均与HCC早期复发、转移密切相关[59],且存在一定的关联性,因此联合预测二者对HCC诊疗具有重要意义。CHU等[60]构建的3D CNN模型显示,多任务学习(同时预测MVI和VETC)模型(AUC=0.917)预测效能显著优于单任务学习(仅预测MVI)模型(AUC=0.896)。QU等[61]亦证实多任务学习模型(AUC=0.855)显著优于单任务学习模型(AUC=0.839)。上述研究均证实多任务学习的优势,其模型对临床治疗方案制定及预后评估具有重要的价值。此外,传统DL模型面临“数据孤岛”的问题,限制了多中心大样本预测模型的训练,而联邦学习[62]作为分布式机器学习,在不收集、不传输原始数据的前提下,实现多参与方的联合模型训练,有望进一步为MVI预测模型优化及临床转化提供更充分的数据支持。

6 不足与展望

       现有研究多为单中心回顾性分析,存在一定局限。一方面数据来源单一、样本量不足,另一方面易产生选择偏倚与信息偏倚,纳排标准难以严格把控,数据质量下降,影响模型的预测性能。其次,数据处理与ROI分割缺乏统一标准,不同研究者、设备及扫描方案对应的勾画尺度、预处理参数、特征提取阈值缺乏统一的行业共识,阻碍了不同研究结果的横向对比。此外,现有DL与影像组学模型多为黑箱模型,可解释性不足,成为制约其临床落地的难题。针对以上问题,未来研究应整合多中心大样本数据,开展多模态前瞻性研究,建立标准化的 ROI分割流程与数据处理体系。在此基础上,引入模型可解释算法,利用注意力热力图、特征贡献分析等手段,解读关键特征对应的生物学意义,打通算法与临床诊断的壁垒,以实现更广泛的临床应用。

7 小结

       综上,不同MRI预测手段评估HCC MVI各有优缺点(表1),GD-EOB-DTPA增强MRI通过其特征性影像学表现,显著提高了对HCC病灶的显示及诊断效能,对HCC术前MVI预测具有极其重要的临床及影像学意义。在此基础上,影像组学、生境成像及DL等技术进一步挖掘瘤内、瘤周及多维度定量特征,有效弥补传统影像主观判读的不足,显著提升MVI预测的准确性与客观性。随着多模态数据融合、模型可解释性优化及外部验证的不断推进,相关研究有望在无创性预测HCC MVI中发挥更大作用,为HCC个体化治疗发挥重要作用。

表1  多种MRI模型预测评估肝细胞癌微血管侵犯的优缺点及展望对比
Tab. 1  Comparison of multiple MRI models for predicting microvascular invasion of hepatocellular carcinoma: merits, drawbacks and prospects

[1]
VOGEL A, MEYER T, SAPISOCHIN G, et al. Hepatocellular carcinoma[J]. Lancet, 2022, 400(10360): 1345-1362. DOI: 10.1016/S0140-6736(22)01200-4.
[2]
LI Z L, XU L D, ZHU S S, et al. Current advances in classification, prediction and management of microvascular invasion in hepatocellular carcinoma[J/OL]. J Cell Mol Med,, 2025, 29(15): e70746 [2026-03-30]. https://doi.org/10.1111/jcmm.70746. DOI: 10.1111/jcmm.70746.
[3]
PEREZ-GIRBES A, LEE J M, MARTÍ-BONMATÍ L. Hepatobiliary contrast agents for liver magnetic resonance imaging[J]. Radiol Engl Ed, 2024, 66: S75-S88. DOI: 10.1016/j.rxeng.2024.05.002.
[4]
YAO W W, ZHANG H W, MA Y P, et al. Comparative analysis of the performance of hepatobiliary agents in depicting MRI features of microvascular infiltration in hepatocellular carcinoma[J]. Abdom Radiol, 2024, 49(7): 2242-2249. DOI: 10.1007/s00261-024-04311-x.
[5]
ZHU Z, WU K Y, LU J, et al. Gd-EOB-DTPA-enhanced MRI radiomics and deep learning models to predict microvascular invasion in hepatocellular carcinoma: a multicenter study[J/OL]. BMC Med Imaging, 2025, 25(1): 105 [2026-03-30]. https://doi.org/10.1186/s12880-025-01646-9. DOI: 10.1186/s12880-025-01646-9.
[6]
HAN X, SHAN L F, XU R, et al. Assessing MRI-based artificial intelligence models for preoperative prediction of microvascular invasion in hepatocellular carcinoma: a systematic review and meta-analysis[J]. Acad Radiol, 2025, 32(11): 6463-6477. DOI: 10.1016/j.acra.2025.06.030.
[7]
YAN M, ZHANG X, ZHANG B, et al. Deep learning nomogram based on Gd-EOB-DTPA MRI for predicting early recurrence in hepatocellular carcinoma after hepatectomy[J]. Eur Radiol, 2023, 33(7): 4949-4961. DOI: 10.1007/s00330-023-09419-0.
[8]
HUANG Z, ZHU R H, LI S S, et al. Comparison of sonazoid-contrast-enhanced ultrasound and Gd-EOB-DTPA-enhanced MRI for predicting microvascular invasion in hepatocellular carcinoma[J]. Ultrasound Med Biol, 2024, 50(9): 1339-1345. DOI: 10.1016/j.ultrasmedbio.2024.05.008.
[9]
HONG S B, CHOI S H, KIM S Y, et al. MRI features for predicting microvascular invasion of hepatocellular carcinoma: A systematic review and meta-analysis[J]. Liver Cancer, 2021, 10(2): 94-106. DOI: 10.1159/000513704.
[10]
LU M T, QU Q, XU L, et al. Prediction for aggressiveness and postoperative recurrence of hepatocellular carcinoma using gadoxetic acid-enhanced magnetic resonance imaging[J]. Acad Radiol, 2023, 30(5): 841-852. DOI: 10.1016/j.acra.2022.12.018.
[11]
KIM K A, KIM M J, JEON H M, et al. Prediction of microvascular invasion of hepatocellular carcinoma: Usefulness of peritumoral hypointensity seen on gadoxetate disodium-enhanced hepatobiliary phase images[J]. Magnetic Resonance Imaging, 2012, 35(3): 629-634. DOI: 10.1002/jmri.22876.
[12]
BEI H Y, KANG Z J, ZHONG M, et al. Combined hematology, tumor morphology, and magnetic resonance imaging for predicting microvascular invasion in hepatocellular carcinoma: a clinical study[J/OL]. Quant Imaging Med Surg, 2026, 16(2): 144 [2026-03-30]. https://doi.org/10.21037/qims-2025-758. DOI: 10.21037/qims-2025-758.
[13]
YANG H L, HAN P, HUANG M T, et al. The role of gadoxetic acid-enhanced MRI features for predicting microvascular invasion in patients with hepatocellular carcinoma[J]. Abdom Radiol (NY), 2022, 47(3): 948-956. DOI: 10.1007/s00261-021-03392-2.
[14]
QIN Y J, ZHANG J, XU D Y, et al. Diagnosis of hepatocellular carcinoma using gadoxetic acid-Enhanced MRI Based on LI-RADS Version 2018 and LI-RADS Modifications[J/OL]. Liver Int, 2025, 45(11): e70366 [2026-03-30]. https://doi.org/10.1111/liv.70366. DOI: 10.1111/liv.70366.
[15]
ZHANG S, XU G Z, DUAN C F, et al. Radiomics analysis of MR imaging with Gd-EOB-DTPA for preoperative prediction of microvascular invasion in hepatocellular carcinoma: investigation and comparison of different hepatobiliary phase delay times[J/OL]. BioMed Res Int, 2021, 2021: 6685723 [2026-03-30]. https://doi.org/10.1155/2021/6685723. DOI: 10.1155/2021/6685723.
[16]
MIN J H, LEE M W, PARK H S, et al. Interobserver variability and diagnostic performance of gadoxetic acid-enhanced MRI for predicting microvascular invasion in hepatocellular carcinoma[J]. Radiology, 2020, 297(3): 573-581. DOI: 10.1148/radiol.2020201940.
[17]
ZHANG K, XIE S S, LI W C, et al. Prediction of microvascular invasion in HCC by a scoring model combining Gd-EOB-DTPA MRI and biochemical indicators[J]. Eur Radiol, 2022, 32(6): 4186-4197. DOI: 10.1007/s00330-021-08502-8.
[18]
FERRARI R, TRINCI M, CASINELLI A, et al. Radiomics in radiology: What the radiologist needs to know about technical aspects and clinical impact[J]. La Radiol Med, 2024, 129(12): 1751-1765. DOI: 10.1007/s11547-024-01904-w.
[19]
XIA T Y, ZHAO B, LI B R, et al. MRI-based radiomics and deep learning in biological characteristics and prognosis of hepatocellular carcinoma: opportunities and challenges[J]. J Magn Reson Imaging, 2024, 59(3): 767-783. DOI: 10.1002/jmri.28982.
[20]
FENG S T, JIA Y M, LIAO B, et al. Preoperative prediction of microvascular invasion in hepatocellular cancer: a radiomics model using Gd-EOB-DTPA-enhanced MRI[J]. Eur Radiol, 2019, 29(9): 4648-4659. DOI: 10.1007/s00330-018-5935-8.
[21]
WANG L L, ZHANG Y Y, LI J F, et al. A nomogram of magnetic resonance imaging for preoperative assessment of microvascular invasion and prognosis of hepatocellular carcinoma[J]. Dig Dis Sci, 2023, 68(12): 4521-4535. DOI: 10.1007/s10620-023-08022-z.
[22]
PENG Y, WU S X, XIONG B, et al. Multiphase MRI radiomics model for predicting microvascular invasion in HCC: Development and clinical validation[J/OL]. iLIVER, 2025, 4(2): 100165 [2026-03-30]. https://doi.org/10.1016/j.iliver.2025.100165. DOI: 10.1016/j.iliver.2025.100165.
[23]
ZENG Y L, WU H Q, ZHU Y Q, et al. MRI-based intra-tumoral ecological diversity features and temporal characteristics for predicting microvascular invasion in hepatocellular carcinoma[J/OL]. Front Oncol, 2025, 15: 1510071 [2026-03-30]. https://doi.org/10.3389/fonc.2025.1510071. DOI: 10.3389/fonc.2025.1510071.
[24]
WANG Q H, ZHOU Y J, YANG H A, et al. MRI-based clinical-radiomics nomogram model for predicting microvascular invasion in hepatocellular carcinoma[J]. Med Phys, 2024, 51(7): 4673-4686. DOI: 10.1002/mp.17087.
[25]
WANG F, YAN C Y, QIN Y, et al. Multiple machine-learning fusion model based on Gd-EOB-DTPA-enhanced MRI and aminotransferase-to-platelet ratio and gamma-glutamyl transferase-to-platelet ratio to predict microvascular invasion in solitary hepatocellular carcinoma: A multicenter study[J]. J Hepatocell Carcinoma, 2024, 11: 427-442. DOI: 10.2147/jhc.s449737.
[26]
ZHANG K, ZHANG L, LI W C, et al. Radiomics nomogram for the prediction of microvascular invasion of HCC and patients' benefit from postoperative adjuvant TACE: a multi-center study[J]. Eur Radiol, 2023, 33(12): 8936-8947. DOI: 10.1007/s00330-023-09824-5.
[27]
PEI J X, WANG L S, LI H. Development of a better nomogram for prediction of preoperative microvascular invasion and postoperative prognosis in hepatocellular carcinoma patients: A comparison study[J]. J Comput Assist Tomogr, 2025, 49(1): 9-22. DOI: 10.1097/RCT.0000000000001618.
[28]
KONG X L, ZHU M T, YE S L, et al. Correlation between 18 f-fdg pet/ct, Gd-eob-dtpa MRI combined with laboratory parameters and microvascular invasion, and the potential of nomogram to predict microvascular invasion of hepatocellular carcinoma before liver transplantation[J]. Nucl Med Commun, 2026, 47(5): 563-571. DOI: 10.1097/MNM.0000000000002117.
[29]
FAYYAZ A M, ABDULKADIR S J, TALPUR N, et al. Grad-CAM (Gradient-weighted Class Activation Mapping): A systematic literature review[J/OL]. Comput Biol Med, 2025, 198: 111200 [2026-03-30]. https://doi.org/10.1016/j.compbiomed.2025.111200. DOI: 10.1016/j.compbiomed.2025.111200.
[30]
WANG J, LIN Z, LIN Z W, et al. Predicting microvascular invasion in solitary hepatocellular carcinoma: a multi-center study integrating clinical, MRI assessments, and radiomics indicators[J/OL]. Front Oncol, 2025, 15: 1511260 [2026-03-30]. https://doi.org/10.3389/fonc.2025.1511260. DOI: 10.3389/fonc.2025.1511260.
[31]
张瑾, 孙中琪, 赵垚, 等. 整合可解释的多区域影像组学和身体成分参数预测肝细胞癌微血管侵犯[J]. 临床放射学杂志, 2026, 45(3): 480-489. DOI: 10.3969/j.issn.1001-9324.2026.03.016.
ZHANG J, SUN Z Q, ZHAO Y, et al. Predicting microvascular invasion in hepatocellular carcinoma with multiregional radiomics and body composition analysis[J]. J Clin Radiol, 2026, 45(3): 480-489. DOI: 10.3969/j.issn.1001-9324.2026.03.016.
[32]
CHEN Y D, XIA Y W, TOLAT P P, et al. Comparison of conventional gadoxetate disodium-enhanced MRI features and radiomics signatures with machine learning for diagnosing microvascular invasion[J]. AJR Am J Roentgenol, 2021, 216(6): 1510-1520. DOI: 10.2214/AJR.20.23255.
[33]
郭剑波, 张添辉, 黄送, 等. 基于钆塞酸二钠增强MRI的机器学习模型可预测肝细胞癌的微血管侵犯[J]. 分子影像学杂志, 2023, 46(4): 736-740. DOI: 10.12122/j.issn.1674-4500.2023.04.28.
GUO J B, ZHANG T H, HUANG S, et al. Machine learning model based on Gd-EOB-DTPA-enhanced MRI in predicting microvascular invasion of hepatocellular carcinoma[J]. J Mol Imaging, 2023, 46(4): 736-740. DOI: 10.12122/j.issn.1674-4500.2023.04.28.
[34]
LI S L, DAI Y M, CHEN J Y, et al. MRI-based habitat imaging in cancer treatment: current technology, applications, and challenges[J/OL]. Cancer Imaging, 2024, 24(1): 107 [2026-03-30]. https://doi.org/10.1186/s40644-024-00758-9. DOI: 10.1186/s40644-024-00758-9.
[35]
LIU Z H, LI X G, HUANG Y, et al. CT-based intratumoral and peritumoral radiomics to predict the treatment response to hepatic arterial infusion chemotherapy plus lenvatinib and PD-1 in high-risk hepatocellular carcinoma cases: a multi-center study[J]. Hepatol Int, 2025, 19(6): 1397-1411. DOI: 10.1007/s12072-025-10877-5.
[36]
ZHU Z W, WU J, GUO Y, et al. Prediction of Ki-67 expression in hepatocellular carcinoma with machine learning models based on intratumoral and peritumoral radiomic features[J/OL]. World J Gastrointest Oncol, 2025, 17(5) [2026-03-30]. https://doi.org/10.4251/wjgo.v17.i5.104172. DOI: 10.4251/wjgo.v17.i5.104172.
[37]
ZHANG Y F, WANG S T, SONG M Y, et al. MRI-based intra- and peritumoral heterogeneity in hepatocellular carcinoma for microvascular invasion prediction and prognostic risk stratification[J/OL]. Radiol Imaging Cancer, 2025, 7(6): e250066 [2026-03-30]. https://doi.org/10.1148/rycan.250066. DOI: 10.1148/rycan.250066.
[38]
蒋韬. 计算机辅助的多序列MRI术前预测肝癌微血管侵犯的研究[D]. 沈阳: 中国医科大学, 2022.
JIANG T. Computer-aided prediction of microvascular invasion in hepatocellular carcinoma based on multiparametric MRI[D]. Shenyang: China Medical University, 2022.
[39]
MIRANDA J, HORVAT N, FONSECA G M, et al. Current status and future perspectives of radiomics in hepatocellular carcinoma[J]. World J Gastroenterol, 2023, 29(1): 43-60. DOI: 10.3748/wjg.v29.i1.43.
[40]
ZHANG Y F, YANG C, QIAN X L, et al. Evaluate the microvascular invasion of hepatocellular carcinoma (≤5 cm) and recurrence free survival with gadoxetate disodium-enhanced MRI-based habitat imaging[J]. J Magn Reson Imaging, 2024, 60(4): 1664-1675. DOI: 10.1002/jmri.29207.
[41]
YU H M, CHEN M, KONG D P, et al. Shap-interpretable predictive modeling of microvascular invasion and early recurrence in hepatocellular carcinoma using MRI habitat imaging combined with clinical features[J/OL]. Eur J Radiol, 2026, 196: 112667 [2026-03-30]. https://doi.org/10.1016/j.ejrad.2026.112667. DOI: 10.1016/j.ejrad.2026.112667.
[42]
CHONG H H, YANG L, SHENG R F, et al. Multi-scale and multi-parametric radiomics of gadoxetate disodium-enhanced MRI predicts microvascular invasion and outcome in patients with solitary hepatocellular carcinoma≤5 cm[J]. Eur Radiol, 2021, 31(7): 4824-4838. DOI: 10.1007/s00330-020-07601-2.
[43]
XIE Y L, ZHANG T, LIU Z X, et al. MRI-based models using habitat imaging for predicting distinct vascular patterns in hepatocellular carcinoma[J]. Acad Radiol, 2025, 32(11): 6491-6502. DOI: 10.1016/j.acra.2025.07.010.
[44]
CHEN M W, YAN Z H, SHI G Z, et al. Preoperative prediction of microvascular invasion in hepatocellular carcinoma using gadoxetic acid-enhanced golden-angle radial sparse parallel dynamic MRI[J/OL]. Eur J Radiol, 2025, 192: 112399 [2026-03-30]. https://doi.org/10.1016/j.ejrad.2025.112399. DOI: 10.1016/j.ejrad.2025.112399.
[45]
MA L D, LIAO S T, YUAN S S, et al. Refining risk stratification in hepatocellular carcinoma by integrating tertiary lymphoid structures and microvascular invasion: a multicenter retrospective study[J]. Int J Surg, 2025, 111(11): 8212-8225. DOI: 10.1097/JS9.0000000000003045.
[46]
SAUTÈS-FRIDMAN C, PETITPREZ F, CALDERARO J, et al. Tertiary lymphoid structures in the era of cancer immunotherapy[J]. Nat Rev Cancer, 2019, 19(6): 307-325. DOI: 10.1038/s41568-019-0144-6.
[47]
李依蔓, 成杰, 陈凤喜, 等. 术前钆塞酸二钠增强磁共振成像检查预测肝细胞癌微血管侵犯和瘤内三级淋巴结构的临床价值[J]. 中华消化外科杂志, 2024, 23(12): 1556-1565. DOI: 10.3760/cma.j.cn115610-20241022-00458.
LI Y M, CHENG J, CHEN F X, et al. Clinical value of preoperative Gd-EOB-DTPA-enhanced magnetic resonance imaging in predic-Ting microvascular invasion and intratumoral tertiary lymphoid structures in hepatocellular carcinoma[J]. Chin J Dig Surg, 2024, 23(12): 1556-1565. DOI: 10.3760/cma.j.cn115610-20241022-00458.
[48]
LUO X, LI P W, CHEN H Y, et al. Automatic segmentation of hepatocellular carcinoma on dynamic contrast-enhanced MRI based on deep learning[J/OL]. Phys Med Biol, 2024, 69(6): 065008 [2026-03-30]. https://doi.org/10.1088/1361-6560/ad2790. DOI: 10.1088/1361-6560/ad2790.
[49]
WEI H, ZHENG T Y, ZHANG X L, et al. Deep learning-based 3D quantitative total tumor burden predicts early recurrence of BCLC A and B HCC after resection[J]. Eur Radiol, 2025, 35(1): 127-139. DOI: 10.1007/s00330-024-10941-y.
[50]
SONG D J, WANG Y Y, WANG W T, et al. Using deep learning to predict microvascular invasion in hepatocellular carcinoma based on dynamic contrast-enhanced MRI combined with clinical parameters[J]. J Cancer Res Clin Oncol, 2021, 147(12): 3757-3767. DOI: 10.1007/s00432-021-03617-3.
[51]
YU H M, KONG D P, MO X J, et al. MRI-based 2.5D deep learning and radiomics effectively predicted microvascular invasion and Ki-67 expression in hepatocellular carcinoma[J/OL]. PLoS One, 2025, 20(11): e0336579 [2026-03-30]. https://doi.org/10.1371/journal.pone.0336579. DOI: 10.1371/journal.pone.0336579.
[52]
YIN Z X, WU M L, LI Y Y, et al. Exploring 2D and 3D radiomic models for predicting microvascular invasion in hepatocellular carcinoma: a novel perspective on tumor heterogeneity[J/OL]. Front Oncol, 2025, 15: 1590655 [2026-03-30]. https://doi.org/10.3389/fonc.2025.1590655. DOI: 10.3389/fonc.2025.1590655.
[53]
WANG T, LI Z, YU H Y, et al. Prediction of microvascular invasion in hepatocellular carcinoma based on preoperative Gd-EOB-DTPA-enhanced MRI: Comparison of predictive performance among 2D, 2D-expansion and 3D deep learning models[J/OL]. Front Oncol, 2023, 13: 987781 [2026-03-30]. https://doi.org/10.3389/fonc.2023.987781. DOI: 10.3389/fonc.2023.987781.
[54]
KUMAR A, JIANG H X, IMRAN M, et al. A flexible 2.5D medical image segmentation approach with in-slice and cross-slice attention[J/OL]. Comput Biol Med, 2024, 182: 109173 [2026-03-30]. https://doi.org/10.1016/j.compbiomed.2024.109173. DOI: 10.1016/j.compbiomed.2024.109173.
[55]
LUO Y, ZHANG G, ZHONG S Y, et al. Deep multi-instance learning model based on gadoxetic acid-enhanced MRI for predicting microvascular invasion of hepatocellular carcinoma: a multicenter, retrospective study[J/OL]. BMC Cancer, 2025, 25(1): 1626 [2026-03-30]. https://doi.org/10.1186/s12885-025-14971-7. DOI: 10.1186/s12885-025-14971-7.
[56]
ELHAIE M, KOOZARI A, ARJMANDI M, et al. Deep learning for hepatocellular carcinoma segmentation in MRI: A systematic review of models, performance, and challenges[J/OL]. Medicine, 2025, 104(51): e47061 [2026-03-30]. https://doi.org/10.1097/md.0000000000047061. DOI: 10.1097/MD.0000000000047061.
[57]
HE R L, CHEN H L, ZOU W J, et al. Transformer-based deep learning for preoperative prediction of microvascular invasion in hepatocellular carcinoma[J/OL]. Cancers, 2025, 17(20): 3314 [2026-03-30]. https://doi.org/10.3390/cancers17203314. DOI: 10.3390/cancers17203314.
[58]
黄倩, 庄银平, 徐鹏, 等. 基于多序列MRI和Swin Transformer深度学习模型的肝细胞癌微血管侵犯预测[J]. 中国医学物理学杂志, 2026, 43(2): 245-254. DOI: 10.3969/j.issn.1005-202X.2026.02.015.
HUANG Q, ZHUANG Y P, XU P, et al. Predicting microvascular invasion in hepatocellular carcinoma with multi-sequence MRI and a Swin Transformer-based deep learning model[J]. Chin J Med Phys, 2026, 43(2): 245-254. DOI: 10.3969/j.issn.1005-202X.2026.02.015.
[59]
WANG M M, WANG Y Z, CAO L, et al. Diagnostic performance based on MRI for preoperative of vessels encapsulating tumor clusters in hepatocellular carcinoma: a systematic review and meta-analysis[J]. La Radiol Med, 2025, 130(12): 1921-1935. DOI: 10.1007/s11547-025-02100-0.
[60]
CHU T J, ZHAO C, ZHANG J, et al. Application of a convolutional neural network for multitask learning to simultaneously predict microvascular invasion and vessels that encapsulate tumor clusters in hepatocellular carcinoma[J]. Ann Surg Oncol, 2022, 29(11): 6774-6783. DOI: 10.1245/s10434-022-12000-6.
[61]
QU Q, LIU Z X, LU M T, et al. Preoperative gadoxetic acid-enhanced MRI features for evaluation of vessels encapsulating tumor clusters and microvascular invasion in hepatocellular carcinoma: creating nomograms for risk assessment[J]. J Magn Reson Imaging, 2024, 60(3): 1094-1110. DOI: 10.1002/jmri.29187.
[62]
LOU L Q, GOVINDARAJAN V, SHAIKH Z A, et al. FedLiverNet: a federated learning framework for privacy-preserving and efficient liver cancer detection[J]. Sci Rep, 2026, 16(1): 16745 [2026-03-30]. https://doi.org/10.1038/s41598-026-46244-z. DOI: 10.1038/s41598-026-46244-z.

上一篇 多参数磁共振成像影像组学在肺癌诊疗中的研究进展
下一篇 基于多参数MRI影像组学与细胞外容积对肝细胞癌微血管侵犯的研究进展
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2