分享:
分享到微信朋友圈
X
临床研究
基于CEMRI瘤内瘤周影像组学预测肝细胞癌分化程度的研究
陆煜杰 顾文豪 许大波 刘海峰 邢伟

Cite this article as: LU Y J, GU W H, XU D B, et al. CEMRI-based intratumoral and peritumoral radiomics for predicting the degree of pathological differentiation of hepatocellular carcinoma[J]. Chin J Magn Reson Imaging, 2025, 16(3): 51-57.本文引用格式:陆煜杰, 顾文豪, 许大波, 等. 基于CEMRI瘤内瘤周影像组学预测肝细胞癌分化程度的研究[J]. 磁共振成像, 2025, 16(3): 51-57. DOI:10.12015/issn.1674-8034.2025.03.008.


[摘要] 目的 建立并验证基于对比增强磁共振成像(contrast enhanced magnetic resonance imaging, CEMRI)瘤内瘤周影像组学模型预测肝细胞癌(hepatocellular carcinoma, HCC)分化程度的价值。材料与方法 回顾性分析2020年1月至2023年7月苏州大学附属第三医院213例经术后病理结果证实为HCC患者的资料(223个病灶),包括62个低度分化的HCC(poorly differentiated HCC, pHCC)、161个非低度分化的HCC(non-poorly differentiated HCC, npHCC)。采用交叉验证方法按照7∶3的比例随机分为训练集(149例,156个HCC病灶)、测试集(64例,67个HCC病灶)。使用ITK-SNAP软件在动脉期、门静脉期和延迟期图像上勾画HCC全域感兴趣区(region of interest, ROI),基于PyRadiomics软件包共提取3045个组学特征,先后采用Spearman相关性分析、最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)回归和最大相关性-最小冗余(maximum relevance-minimum redundancy, mRMR)方法进行数据降维并选择最佳特征,随后使用支持向量机算法分别构建瘤内(Intratumoral)、瘤周5 mm(Peri_5mm)、瘤周10 mm(Peri_10mm)模型,并融合瘤内及最佳瘤周参数构建瘤内瘤周融合(IntraPeri)模型。基于受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)和决策曲线分析(decision curve analysis, DCA)评估影像组学模型预测pHCC效能及临床效益。结果 Intratumoral、Peri_5mm、Peri_10mm、IntraPeri模型分别纳入10、17、11、12个组学特征。Intratumoral模型预测pHCC训练集和测试集的AUC分别为0.92、0.93;Peri_10mm模型预测pHCC的AUC值在训练集(0.88 vs. 0.82)、测试集(0.90 vs. 0.85)均高于Peri_5mm模型。IntraPeri模型预测pHCC效能最佳,在训练集和测试集AUC值分别为0.95、0.95。DCA提示Intratumoral模型及Peri_10mm模型均具有良好的临床收益,其中IntraPeri模型最佳。结论 基于CEMRI的瘤内瘤周影像组学模型可准确预测HCC分化程度,并且具有较好临床受益。
[Abstract] Objective To develop and validate intratumoral and multiregion peritumoral radiomics models based on contrast-enhanced magnetic resonance imaging (CEMRI) for predicting pathological differentiation in hepatocellular carcinoma (HCC) patients.Materials and Methods A total of 213 HCC patients diagnosed between January 2020 and July 2023 at the Third Affiliated Hospital of Soochow University was included in the retrospective study, comprising 62 poorly differentiated HCC (pHCC) and 161 non-poorly differentiated HCCs (npHCC). The HCCs were randomly divided into training (149 patients, 156 HCCs) and validation (64 patients, 67 HCCs) cohorts at a 7∶3 ratio. The ITK-SNAP software delineated the region of interest (ROI) on arterial, portal vein, and delayed phase images, while PyRadiomics software extracted 3045 radiomic features. Feature selection was carried out using Spearman rank correlation, least absolute shrinkage and selection operator (LASSO), and maximum relevance-minimum redundancy (mRMR) approaches, followed by support vector machine algorithm to build Intratumoral, 5 mm peritumoral (Peri_5mm), 10 mm peritumoral (Peri_10mm), and Intratumoral + 10 mm peritumoral (IntraPeri) models. The predictive performance of these models was assessed using the area under the curve (AUC) of receiver operating characteristic and decision curve analysis (DCA).Results The Intratumoral, Peri_5mm, Peri_10mm, and IntraPeri models consisted of 10, 17, 11, and 12 features, respectively. In the Intratumoral model, the AUC values for predicting pHCC in the training and validation cohorts were 0.92 and 0.93, respectively. The Peri_10mm model exhibited higher AUCs compared to the Peri_5mm model: 0.88 versus 0.82 in the training cohort and 0.90 versus 0.85 in the validation cohort. The IntraPeri model demonstrated superior performance with AUC values of 0.95 and 0.95 in the training and validation cohorts, respectively. DCA suggested that the Intratumoral, Peri_5mm, and Peri_10mm models provided notable clinical benefits, with the IntraPeri model being the most optimal.Conclusions The IntraPeri model based on CEMRI can accurately predict HCC differentiation and has good clinical benefits.
[关键词] 肝细胞癌;分化程度;磁共振成像;瘤内;瘤周;影像组学
[Keywords] hepatocellular carcinoma;pathological differentiation;magnetic resonance imaging;intratumoral;peritumoral;radiomics

陆煜杰 1, 2   顾文豪 2   许大波 2   刘海峰 3   邢伟 1, 3*  

1 扬州大学医学院,扬州 225009

2 太仓市第一人民医院影像科,苏州 215400

3 苏州大学附属第三医院放射科,常州 213000

通信作者:邢伟,E-mail: suzhxingwei@suda.edu.cn

作者贡献声明:邢伟设计本研究的方案,对稿件重要内容进行了修改;陆煜杰起草和撰写稿件,获取、分析或解释本研究的数据;顾文豪、许大波、刘海峰获取、分析或解释本研究的数据,对稿件重要内容进行了修改;顾文豪获得了2024年度苏州市应用基础研究(医疗卫生)科技创新(第二批)指导性项目资助;刘海峰获得了2024年常州市第一人民医院临床研究专项资助;全体作者都同意发表最后的修改稿,同意对本研究的所有方面负责,确保本研究的准确性和诚信。


基金项目: 2024年常州市第一人民医院临床研究专项 2024-14 2024年度苏州市应用基础研究(医疗卫生)科技创新(第二批)指导性项目 SYWD2024020
收稿日期:2024-08-31
接受日期:2025-02-27
中图分类号:R445.2  R735.7 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2025.03.008
本文引用格式:陆煜杰, 顾文豪, 许大波, 等. 基于CEMRI瘤内瘤周影像组学预测肝细胞癌分化程度的研究[J]. 磁共振成像, 2025, 16(3): 51-57. DOI:10.12015/issn.1674-8034.2025.03.008.

0 引言

       肝癌是高发病率的常见恶性肿瘤,居我国恶性肿瘤发病率第5位、病死率第2位[1],肝细胞癌(hepatocellular carcinoma, HCC)是最常见的原发性肝癌病理类型(75%~85%)。HCC病理分化程度反映肿瘤异质性及生物学特征,与患者术后高复发率(50%~70%)及生存预后差密切相关[2, 3, 4, 5]。与高分化HCC(well differentiated HCC)及中分化HCC(moderately differentiated HCC)相比,低分化HCC(poorly differentiated HCC, pHCC)术后复发率更高、总生存期更低。因此我们把中分化HCC和高分化HCC统称为非低分化HCC(Non-poorly differentiated HCC, npHCC)[6, 7]。大多数患者初诊时已是晚期,故5年生存率<10%[8],且肝部分切除术后辅助治疗对减少pHCC术后复发有重要的临床获益[9]

       目前肝脏穿刺活检是术前诊断HCC分化程度的参考标准,但穿刺活检存在有创性、采样误差、肿瘤破裂、出血及针道转移等缺陷不适用于常规临床诊断 [6, 10, 11]。因此,术前无创预测HCC分化程度对患者治疗、预后有重要临床意义。

       影像组学能从图像中提取大量肉眼无法识别的高通量特征,定量分析肿瘤异质性及生物学特征,在预测HCC病理表征、疗效监测、生存预后方面有良好技术优势及应用前景[12]。增强磁共振成像(contrast-enhanced magnetic resonance imaging, CEMRI)软组织分辨率及空间分辨率高,能准确揭示HCC异质性[13],是诊断及鉴别诊断HCC的优选影像学方法。

       当前影像组学研究仅探索了MRI瘤内影像组学模型预测HCC分化的价值[14, 15, 16, 17],表明了瘤内影像组学能准确预测HCC病理及生物学特征。但目前越来越多的研究表明瘤周区域的异质性与HCC病理生物特征、疗效及生存预后密切相关[18, 19, 20, 21],且融合瘤内瘤周区域特征的影像组学模型有助于提高预测HCC生物学特征的效能。上述研究主要探究了瘤周5、10及20 mm与瘤内区域的预测效能的差异,却未探究不同范围瘤周区域之间预测效能的差异。更重要的是,目前鲜有CEMRI瘤内瘤周影像组学模型预测HCC病理分化的研究。有研究表明,对比单期相模型,多期相的融合模型在HCC早期诊断[22]、微血管侵犯预测[23, 24]、病理分级预测[17]、预后[25]等方面均表现出更佳的效能。

       因此,本研究旨在构建基于CEMRI完整序列的瘤内及不同瘤周区域影像组学模型,评估并找出预测HCC分化程度效能最佳的模型,以媲美术前穿刺活检的效果,减少穿刺活检的情况,避免并发症,为HCC患者术前诊疗及生存预后提供无创生物影像学依据。

1 材料与方法

1.1 研究对象

       本研究严格遵循《赫尔辛基宣言》,经苏州大学附属第三医院伦理委员会审核批准通过,免除受试者知情同意,伦理批准文号:(2022)科第027号。回顾性分析2020年1月至2023年7月于苏州大学附属第三医院诊断HCC患者的资料。纳入标准:(1)术前进行上腹部CEMRI检查;(2)经肝脏穿刺活检和/或肝部分切除术后病理证实为HCC。排除标准:(1)CEMRI和手术间隔超过1个月;(2)CEMRI伪影重导致图像质量较差;(3)既往HCC切除、介入和/或免疫治疗病史;(4)HCC病理报告缺失或不完整,不能明确HCC分化程度。本研究初步收集235例患者,因术前治疗史排除5例、CEMRI检查与病理结果间隔超过1月排除5例、图像质量较差排除6例、病理报告不完整排除6例。

1.2 患者基线资料

       本研究最终纳入的213例HCC患者,其中10例患者2个HCC病灶,故最终包含223个HCC病灶,基于HCC数量采用交叉验证方法按照7∶3的比例随机分为训练集(149例,156个HCC病灶)、测试集(64例,67个HCC病灶)。从医院电子病历系统收集患者基线资料:性别、年龄、病因、甲胎蛋白、谷丙转氨酶、谷草转氨酶、总胆红素、凝血时间、白蛋白、Child-Pugh分级、巴塞罗那(Barcelona clinical liver cancer, BCLC)分期。

1.3 CEMRI扫描方案

       采用西门子Verio 3.0 T扫描仪、12通道腹部矩阵线圈完成上腹部CEMRI扫描。扫描参数:使用3D容积插值屏气检查技术,TR 3.92 ms,TE 1.39 ms,对比剂采用钆喷酸葡胺(北京北陆药业股份有限公司,中国),剂量0.2 mL/kg,流率1.0 mL/s,注射后25~35 s、60~70 s、180 s采集动脉期(arterial phase, AP)、门静脉期(portal veinous phase, PVP)和延迟期(delayed phase, DP)图像。

1.4 病理分析

       切除的HCC标本经过HE染色处理后,由一位有15以上年丰富经验的病理医师在不知临床资料和影像学诊断的前提下评定HCC分化程度。根据WHO分类标准,HCC分类为高度分化、中度、低度分化的HCC(图1)。当HCC显示出不同的分化结果时,主导的分化结果确定最终病理分化诊断结果。本研究将中度和高度分化的HCC都归类为npHCC。本研究包括62个pHCC和161个npHCC。

图1  不同分化的肝细胞癌(HCC)病理图(HE ×20)。1A:高分化HCC,癌细胞排列成1~3 层细胞厚度的梁索状(黑箭);1B:中分化HCC,癌细胞梁索的厚度在4~10 层,核/质比例增大,核异型明显(黑箭);1C:低分化HCC,癌细胞呈片状或实性弥漫排列,核异型(白箭)与核分裂(黑箭)明显。
Fig. 1  Pathological picture (HE × 20) of different differentiation of hepatocellular carcinoma (HCC). 1A: Well differentiated HCC, where cancer cells arrange in trabecular structures with 1 to 3 layers thick (black arrow); 1B: Moderately differentiated HCC, where cells arrange in trabecular structures with 4 to 10 layers, with an increased nuclear/cytoplasmic ratio and obvious nuclear atypia (black arrow); 1C: Poorly differentiated HCC, where cancer cells are arranged in sheets or solid diffuse patterns, with noticeable nuclear atypia and mitosis.

1.5 勾画瘤内瘤周感兴趣区

       勾画感兴趣区(region of interest, ROI)之前,CEMRI图像分别进行体素间距重采样为1 mm×1 mm×1 mm以补偿体素空间差异,灰度归一化保持灰度一致性。放射医师1(主治医师,5年工作经验)基于开源软件ITK-SNAP(版本3.6.0,www.itk-snap.org)在动脉期、门静脉期和延迟期图像逐层连续勾画HCC区域,注意避开卫星结节及瘤周灌注异常区域,软件自动生成HCC全域3D ROI。放射医师2(主治医师,8年工作经验)校准ROI区域,如有分歧则通过讨论解决。

       基于Python软件(版本3.6)中的SimpleITK软件包自动外扩瘤内区域生成瘤周区域。由于CEMRI图像体素间距已被重采样为1 mm×1 mm×1 mm,因此扩张后的体素(5、10)通过3D盒状核进行卷积扩张成相应的瘤周区域5 mm、10 mm,超出肝实质范围的区域由放射医师人工校正(图2)。

图2  基于对比增强磁共振成像(CEMRI)门脉期勾画的肿瘤(红色)及外扩后的瘤周5 mm、10 mm(绿色)感兴趣区。
Fig. 2  Examples of ROI based on portal veinous phase of contrast enhanced magnetic resonance imaging (CEMRI). The red and green region represent the intratumoral region and the multiple peritumoral regions respectively.

1.6 影像组学特征提取、降维及模型构建

1.6.1 影像组学特征提取

       使用Python软件PyRadiomics程序包提取HCC全域形状学特征(n=14)、一阶直方图特征(n=198)和纹理分析特征(n=803)。因此,每个HCC病灶在动脉期、门静脉期和延迟期序列图像均分别提取1015个特征,故基于CEMRI每个HCC病灶在瘤内(Intratumoral)、瘤周5 mm(Peri_5mm)、瘤周10 mm(Peri_10mm)区域均分别累计提取3045个特征。

1.6.2 影像组学特征数值标准化

       影像组学特征数值在数量级或者量纲上存在较大的差异,因此对提取的影像组学特征采用Z分数(Z-score)进行标准化处理,使其满足标准正态分布(均值为0,标准差为1),以保证实验结果的可靠性。

1.6.3 影像组学特征一致性评估

       勾画完ROI两周后,随机挑选30例患者的CEMRI图像,由放射医师1和放射医师2重新勾画ROI并提取影像组特征,采用组内相关系数(intra-class correlation coefficient, ICC)对提取的Intratumoral特征进行组内和组间一致性评估。

1.6.4 影像组学特征降维及筛选

       首先对提取的组学特征进行Spearman相关性分析去除冗余,相关系数大于0.9的特征两者保留其一;随后结合十折交叉验证与最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)方法,根据最优惩罚参数选择具有系数不为零的特征;最后进行最大相关性-最小冗余(maximum relevance-minimum redundancy, mRMR)方法进一步进行数据降维选择最佳特征。

1.6.5 影像组学模型构建

       将最后选择的最佳影像组学特征依据各自系数组化起来并输入到支持向量机(support vector machine, SVM)分类器,分别构建Intratumoral、Peri_5mm、Peri_10mm模型预测pHCC;同时融合最佳瘤周区域和瘤内影像组学特征构建瘤内瘤周融合(IntraPeri)模型。

1.7 统计学分析

       使用R软件(版本4.0)进行统计分析。分类数据以百分比形式呈现,而连续变量经过Shapiro-Wilk方法进行正态性检验后,符合正态分布以平均数±标准差的形式表达,否则使用中位数(四分位距)的形式表达。使用Mann-Whitney U检验或卡方(χ2)检验比较训练集和测试集临床基线资料差异。ICC值≥0.75表示信度较好,特征具有较好的一致性。采用受试者工作特征(receiver operating characteristic curve, ROC)曲线下面积(area under the curve, AUC)评估各影像组学模型预测pHCC的效能,并采用决策曲线分析(decision curve analysis, DCA)评估临床净收益,最后用DeLong方法比较各模型AUC值之间的差异。

2 结果

2.1 临床基线资料

       本研究共纳入213例患者,其中10例患者为2个HCC病灶,共223个病灶。按7∶3的比例随机分为训练集(149例,156个病灶)、测试集(64例,67个病灶)。训练集和测试集HCC患者的年龄、性别、病因、甲胎蛋白、谷丙转氨酶、谷草转氨酶、总胆红素、凝血时间、白蛋白、Child-Pugh分级、HCC数量、BCLC分期差异均无统计学意义(P>0.05),详见表1所示。

表1  纳入HCC患者的临床基线资料
Tab. 1  Baseline characteristics of included patients diagnosed with HCC

2.2 影像组学特征一致性评估

       基于随机抽取的30例图像提取Intratumoral影像组学特征,组内和组间ICC分别为0.89、0.87,证实本研究提取的影像组学特征具有较好的稳定性及一致性。

2.3 影像组学特征提取与筛选

       将提取的3045个特征通过Spearman相关性分析去除冗余特征后,分别保留了Intratumoral、Peri_5mm、Peri_10mm、IntraPeri模型的298、279、292、265个影像组学特征;然后使用LASSO及mRMR算法进一步对数据降维,选择出最佳影像组学特征,包括10个Intratumoral特征、17个Peri_5mm特征、11个Peri_10mm特征、12个IntraPeri特征,分别构建Intratumoral、Peri_5mm、Peri_10mm及IntraPeri影像组学模型(图3)。

图3  IntraPeri影像组学特征提取流程。3A:LASSO回归纵轴表示3045个特征对应的系数,横轴表示调整参数λ;3B:通过10折mRMR交叉验证,选择最佳λ(λ=0.045)值;3C:提取的12个最佳IntraPeri影像组学特征(系数不为0)。IntraPeri:融合瘤内及瘤周模型构建的瘤内瘤周融合模型;LASSO:最小绝对收缩和选择算子;mRMR:最大相关性-最小冗余。
Fig. 3  Process of extracting radiomics features. 3A: The vertical axis of LASSO regression represents the coefficients corresponding to 3045 features, and the horizontal axis represents the adjustment parameter λ; 3B: The optimal λ (λ=0.045) is selected through mRMR; 3C: 12 IntraPeri optimal features have been extracted. IntraPeri: Intratumoral + 10 mm peritumoral combined model; LASSO: least absolute shrinkage and selection operator; mRMR: maximum relevance-minimum redundancy.

2.4 瘤内、瘤周及融合影像组学模型诊断效能

       Intratumoral模型预测pHCC的AUC值在训练集、测试集分别为0.92(95% CI:0.87~0.97)、0.93(95% CI:0.88~0.97),高于Peri_10mm模型[0.88(95% CI:0.83~0.93)、0.90(95% CI:0.83~0.96)],Peri_5mm模型AUC值最低[0.82(95% CI:0.76~0.88)、0.85(95% CI:0.78~0.92)],如表2所示。通过融合瘤内和瘤周10 mm影像组学特征构建IntraPeri模型预测pHCC的AUC值在训练集和测试集分别为0.95(95% CI:0.91~0.98)、0.95(95% CI:0.92~0.99)。

表2  不同瘤内瘤周影像组学模型诊断pHCC的效能
Tab. 2  Diagnostic performance of the different model for predicting pHCC

2.5 DCA及DeLong检验结果

       Intratumoral、Peri_5mm、Peri_10mm、IntraPeri模型的DCA结果在所有阈值范围内位于Treat ALL线、Treat None线上方,提示所有影像组学模型均具有临床受益。DeLong检验提示Intratumoral、Peri_10mm模型预测pHCC在训练集的AUC值均高于Peri_5mm(P<0.05),但Intratumoral、Peri_10mm模型间诊断效能差异无统计学意义(P=0.140);IntraPeri模型预测pHCC效能高于Peri_5mm、Peri_10mm组学模型(P<0.05),但与Intratumoral模型诊断效能差异无统计学意义(P=0.172)(图4)。

图4  不同影像组学模型在训练集和测试集预测pHCC的ROC曲线、DCA及DeLong检验结果。pHCC:低度分化的肝细胞癌;ROC:受试者工作特征;DCA:决策曲线分析。
Fig. 4  The ROC curve, the result of DCA and DeLong test of each model in predicting pHCC in the training and validation cohorts. pHCC: poorly differentiated hepatocellular carcinoma; ROC: receiver operating characteristic; DCA: decision curve analysis.

3 讨论

       本研究较先证实基于CEMRI的瘤内、瘤周影像组学模型对预测pHCC有较高准确性,尤其是瘤内模型对诊断pHCC效能极佳,训练集和测试集的AUC值分别为0.92、0.93;同时,通过融合瘤内和瘤周10 mm组学特征构建的IntraPeri模型能进一步提高预测pHCC的效能,训练集和测试集的AUC值分别为0.95、0.95;且DCA结果证实CEMRI瘤内瘤周影像组学模型均具有较高临床净受益,有助于HCC术前无创精准判断病理分化程度,从而为治疗方式选择及改善预后提供客观有效生物影像学依据。

3.1 瘤内影像组学模型诊断pHCC效能

       常规MRI语义特征如肿瘤大小、边界不清、马赛克征及晕环状强化与HCC分化程度密切相关[13],有助于术前鉴别诊断pHCC。LIU等[14]联合MRI形态学特征预测pHCC的AUC为0.77(95% CI:0.70~0.83),寻求更加精准预测HCC分化程度的研究方法成为当前研究热点。影像组学是一种新兴的强大的非侵入性工具,可以提取人眼无法识别的高通量影像特征,能全面反映肿瘤异质性,在HCC术前诊断、病理特征预测及预后判断等方面有良好技术优势及应用前景[26, 27, 28, 29]。近期YANG等[30]和BRANCATO等[31]都构建了基于CEMRI的瘤内影像组学模型预测HCC病理分化程度,但纳入的病例数目有限(38~188例)、提取的影像组学特征较少(38~108个),这可能是导致研究中AUC值较低(0.58~0.74)的主要原因;同时两项研究仅仅提取分析了瘤内组学特征,并未探究瘤周异质性对诊断HCC分化程度的影响。本研究基于213例患者的223个HCC病灶构建瘤内模型对预测pHCC的AUC值较高,不仅仅与本研究基于CEMRI序列提取的3045个组学特征能充分反映HCC异质性程度有关[30, 31, 32];还与我们依次使用Spearman相关性分析、LASSO算法及mRMR算法进行降维确定最佳特征、减少特征过拟合有关,同时使用SVM算法分类器构建模型解决数据维度高的问题[33, 34]

3.2 瘤周影像组学模型诊断pHCC效能及差异分析

       瘤周区域是最早且最常易受肿瘤异质性影响的组织,进而成为门静脉肿瘤血栓形成和转移的主要血行播散途径[35]。瘤周区域主要包括汇管区炎症(主要是淋巴细胞浸润)、异常灌注、肝窦扩张、微血管侵犯等,是反映HCC异质性及生物学特征的关键组成部分[36]。本研究分别构建瘤周5 mm和瘤周10 mm影像组学模型,证实瘤周区域异质性对预测pHCC有重要的补充价值。JI等[37]和WANG等[38]也证实瘤周区域特征对预测HCC术后复发具有重要价值。同时值得注意的是,Peri_10mm模型诊断效能高于Peri_5mm模型,提示不同范围瘤周区域异质性与HCC分化程度密切相关。Li等[18]的研究表明瘤周0~10 mm范围代表HCC的微转移区域,与HCC复发有明显相关性,特别是晚期复发。CHONG等[24]发现瘤周10 mm影像组学模型预测HCC微血管侵犯准确性高于瘤周5 mm模型,推测与瘤周5 mm区域的范围太窄、缺乏足够的信息,不能完全反映HCC异质性有关。

       更要的是,本研究还发现瘤内影像组学模型预测pHCC的效能优于瘤周影像组学模型,这主要由HCC分化病理诊断所决定:HCC内部癌细胞的大小、异质性程度、病理核分裂象及计数等是病理诊断HCC分化的关键依据[39],因此瘤内模型较瘤周模型更能反映HCC分化程度。

3.3 IntraPeri模型及临床应用评估

       IntraPeri模型对提高预测效能有重要作用。本研究通过融合瘤内、瘤周10 mm影像组学构建的IntraPeri模型能提高预测pHCC的效能,训练集和测试集AUC值均高达0.95。因此,我们推测瘤内特征及瘤周特征可以形成互补为预测HCC分化程度提供更高的准确度,且DCA结果证实IntraPeri模型具有临床净收益。近期WANG等[40]和ZHAO等[41]也证实了融合了瘤内瘤周特征的模型对于HCC的诊断和治疗效果预测有更佳的效能。因此,本研究有助于拓展影像组学预测HCC分化的思路,为患者的个性化临床诊疗方案提供指导、选择最佳诊疗方案,改善患者预后。

3.4 本研究的局限性

       (1)本研究是一项回顾性研究,排除了病理报告缺乏及图像质量差的HCC病例,潜在诱发选择偏倚;纳入研究数量较少且并未进行外部验证。后续我们将进行多中心大样本的前瞻性研究验证研究结果的临床适用性。(2)CEMRI能更加准确反映HCC生物学特征,故本研究仅提取CEMRI的影像组学特征,后续将进一步探讨T2WI及DWI组学模型的效能。(3)本研究仅探索了影像组学模型与HCC分化程度预测的相关性,影像组学特征与HCC复发和预后的研究对提高HCC诊疗具有重要价值。

4 结论

       综上所述,本研究基于CEMRI,采用影像组学与机器学习方法,筛选出了最佳影像组学特征,构建的瘤内瘤周模型在预测HCC分化程度方面具有较高效能,可以媲美穿刺活检的效果,减少穿刺活检的情况,减轻患者痛苦,避免并发症,为HCC术前无创精准诊断、病理分化预测、个性化诊疗方案的选择及改善预后提供有效生物影像学依据。

[1]
HAN B F, ZHENG R S, ZENG H M, et al. Cancer incidence and mortality in China, 2022[J]. J Natl Cancer Cent, 2024, 4(1): 47-53. DOI: 10.1016/j.jncc.2024.01.006.
[2]
LIU H F, WANG M, WANG Q, et al. Multiparametric MRI-based intratumoral and peritumoral radiomics for predicting the pathological differentiation of hepatocellular carcinoma[J/OL]. Insights Imaging, 2024, 15(1): 97 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/38536542/. DOI: 10.1186/s13244-024-01623-w.
[3]
MO Z Y, CHEN P Y, LIN J, et al. Pre-operative MRI features predict early post-operative recurrence of hepatocellular carcinoma with different degrees of pathological differentiation[J]. Radiol Med, 2023, 128(3): 261-273. DOI: 10.1007/s11547-023-01601-0.
[4]
SHINKAWA H, TANAKA S, KABATA D, et al. The prognostic impact of tumor differentiation on recurrence and survival after resection of hepatocellular carcinoma is dependent on tumor size[J]. Liver Cancer, 2021, 10(5): 461-472. DOI: 10.1159/000517992.
[5]
SINGAL A G, LLOVET J M, YARCHOAN M, et al. AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma[J]. Hepatology, 2023, 78(6): 1922-1965. DOI: 10.1097/HEP.0000000000000466.
[6]
European Association for the Study of the Liver. EASL Clinical Practice Guidelines on the management of hepatocellular carcinoma[J]. J Hepatol, 2025, 82(2): 315-374. DOI: 10.1016/j.jhep.2024.08.028.
[7]
CALDERARO J, ZIOL M, PARADIS V, et al. Molecular and histological correlations in liver cancer[J]. J Hepatol, 2019, 71(3): 616-630. DOI: 10.1016/j.jhep.2019.06.001.
[8]
PARK I, LEE H B, KIM N, et al. Uncovering gene expression signatures and diagnostic–Biomarkers in hepatocellular carcinoma through multinomial logistic regression analysis[J]. J Biotechnol, 2024, 395: 31-43. DOI: 10.1016/j.jbiotec.2024.09.003.
[9]
DU J S, HSU S H, WANG S N. The current and prospective adjuvant therapies for hepatocellular carcinoma[J/OL]. Cancers, 2024, 16(7): 1422 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/38611100/. DOI: 10.3390/cancers16071422.
[10]
KLEINER D E. Hepatocellular carcinoma: liver biopsy in the balance[J]. Hepatology, 2018, 68(1): 13-15. DOI: 10.1002/hep.29831.
[11]
LEHRICH B M, ZHANG J, MONGA S P, et al. Battle of the biopsies: Role of tissue and liquid biopsy in hepatocellular carcinoma[J]. J Hepatol, 2024, 80(3): 515-530. DOI: 10.1016/j.jhep.2023.11.030.
[12]
XIA T Y, ZHOU Z H, MENG X P, et al. Predicting microvascular invasion in hepatocellular carcinoma using CT-based radiomics model[J/OL]. Radiology, 2023, 307(4): e222729 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/37097141/. DOI: 10.1148/radiol.222729.
[13]
CANDITA G, ROSSI S, CWIKLINSKA K, et al. Imaging diagnosis of hepatocellular carcinoma: a state-of-the-art review[J/OL]. Diagnostics, 2023, 13(4): 625 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/36832113/. DOI: 10.3390/diagnostics13040625.
[14]
LIU H F, LU Y, WANG Q, et al. Machine learning-based CEMRI radiomics integrating LI-RADS features achieves optimal evaluation of hepatocellular carcinoma differentiation[J]. J Hepatocell Carcinoma, 2023, 10: 2103-2115. DOI: 10.2147/JHC.S434895.
[15]
AMELI S, VENKATESH B A, SHAGHAGHI M, et al. Role of MRI-derived radiomics features in determining degree of tumor differentiation of hepatocellular carcinoma[J/OL]. Diagnostics, 2022, 12(10): 2386 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/36292074/. DOI: 10.3390/diagnostics12102386.
[16]
MAO Y F, WANG J C, ZHU Y, et al. Gd-EOB-DTPA-enhanced MRI radiomic features for predicting histological grade of hepatocellular carcinoma[J]. Hepatobiliary Surg Nutr, 2022, 11(1): 13-24. DOI: 10.21037/hbsn-19-870.
[17]
YANG Y, ZHANG S, CUI C, et al. Multiphase MRI-based radiomics for predicting histological grade of hepatocellular carcinoma[J]. J Magn Reson Imaging, 2024, 60(5): 2117-2127. DOI: 10.1002/jmri.29289.
[18]
LI N, WAN X T, ZHANG H, et al. Tumor and peritumor radiomics analysis based on contrast-enhanced CT for predicting early and late recurrence of hepatocellular carcinoma after liver resection[J/OL]. BMC Cancer, 2022, 22(1): 664 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/35715783/. DOI: 10.1186/s12885-022-09743-6.
[19]
HU F, ZHANG Y H, LI M, et al. Preoperative prediction of microvascular invasion risk grades in hepatocellular carcinoma based on tumor and peritumor dual-region radiomics signatures[J/OL]. Front Oncol, 2022, 12: 853336 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/35392229/. DOI: 10.3389/fonc.2022.853336.
[20]
GAO L M, XIONG M L, CHEN X J, et al. Multi-region radiomic analysis based on multi-sequence MRI can preoperatively predict microvascular invasion in hepatocellular carcinoma[J/OL]. Front Oncol, 2022, 12: 818681 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/35574328/. DOI: 10.3389/fonc.2022.818681.
[21]
HE H L, WANG Q, LIU L, et al. Peritumoral edema in preoperative magnetic resonance imaging is an independent prognostic factor for hepatocellular carcinoma[J]. Clin Imaging, 2021, 75: 143-149. DOI: 10.1016/j.clinimag.2021.01.022.
[22]
SUN K, SHI L T, QIU J F, et al. Multi-phase contrast-enhanced magnetic resonance image-based radiomics-combined machine learning reveals microscopic ultra-early hepatocellular carcinoma lesions[J]. Eur J Nucl Med Mol Imaging, 2022, 49(8): 2917-2928. DOI: 10.1007/s00259-022-05742-8.
[23]
YANG Y, FAN W J, GU T, et al. Radiomic features of multi-ROI and multi-phase MRI for the prediction of microvascular invasion in solitary hepatocellular carcinoma[J/OL]. Front Oncol, 2021, 11: 756216 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/34692547/. DOI: 10.3389/fonc.2021.756216.
[24]
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.
[25]
ZHAO Y, WU J J, ZHANG Q H, et al. Radiomics analysis based on multiparametric MRI for predicting early recurrence in hepatocellular carcinoma after partial hepatectomy[J]. J Magn Reson Imaging, 2021, 53(4): 1066-1079. DOI: 10.1002/jmri.27424.
[26]
XU X, ZHANG H L, LIU Q P, et al. Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma[J]. J Hepatol, 2019, 70(6): 1133-1144. DOI: 10.1016/j.jhep.2019.02.023.
[27]
MOKRANE F Z, LU L, VAVASSEUR A, et al. Radiomics machine-learning signature for diagnosis of hepatocellular carcinoma in cirrhotic patients with indeterminate liver nodules[J]. Eur Radiol, 2020, 30(1): 558-570. DOI: 10.1007/s00330-019-06347-w.
[28]
LIU N, WU Y K, TAO Y Y, et al. Differentiation of hepatocellular carcinoma from intrahepatic cholangiocarcinoma through MRI radiomics[J/OL]. Cancers, 2023, 15(22): 5373 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/38001633/. DOI: 10.3390/cancers15225373.
[29]
HECTORS S J, LEWIS S, BESA C, et al. MRI radiomics features predict immuno-oncological characteristics of hepatocellular carcinoma[J]. Eur Radiol, 2020, 30(7): 3759-3769. DOI: 10.1007/s00330-020-06675-2.
[30]
YANG X Z, YUAN C W, ZHANG Y H, et al. Magnetic resonance radiomics signatures for predicting poorly differentiated hepatocellular carcinoma: A SQUIRE-compliant study[J/OL]. Medicine, 2021, 100(19): e25838 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/34106622/. DOI: 10.1097/MD.0000000000025838.
[31]
BRANCATO V, GARBINO N, SALVATORE M, et al. MRI-based radiomic features help identify lesions and predict histopathological grade of hepatocellular carcinoma[J/OL]. Diagnostics, 2022, 12(5): 1085 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/35626241/. DOI: 10.3390/diagnostics12051085.
[32]
KOCAK B, BAESSLER B, BAKAS S, et al. CheckList for EvaluAtion of radiomics research (CLEAR): a step-by-step reporting guideline for authors and reviewers endorsed by ESR and EuSoMII[J/OL]. Insights Imaging, 2023, 14(1): 75 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/37142815/. DOI: 10.1186/s13244-023-01415-8.
[33]
MUKHERJEE S, PATRA A, KHASAWNEH H, et al. Radiomics-based machine-learning models can detect pancreatic cancer on prediagnostic computed tomography scans at a substantial lead time before clinical diagnosis[J]. Gastroenterology, 2022, 163(5): 1435-1446. DOI: 10.1053/j.gastro.2022.06.066.
[34]
ZHENG Y L, ZHOU D, LIU H, et al. CT-based radiomics analysis of different machine learning models for differentiating benign and malignant parotid tumors[J]. Eur Radiol, 2022, 32(10): 6953-6964. DOI: 10.1007/s00330-022-08830-3.
[35]
YANG J W, DONG X, JIN S Z, et al. Radiomics model of dynamic contrast-enhanced MRI for evaluating vessels encapsulating tumor clusters and microvascular invasion in hepatocellular carcinoma[J]. Acad Radiol, 2025, 32(1): 146-156. DOI: 10.1016/j.acra.2024.07.007.
[36]
ZHANG T C, LEI X J, JIA W L, et al. Peritumor tertiary lymphoid structures are associated with infiltrating neutrophils and inferior prognosis in hepatocellular carcinoma[J]. Cancer Med, 2023, 12(3): 3068-3078. DOI: 10.1002/cam4.5227.
[37]
JI G W, ZHU F P, XU Q, et al. Radiomic features at contrast-enhanced CT predict recurrence in early stage hepatocellular carcinoma: a multi-institutional study[J]. Radiology, 2020, 294(3): 568-579. DOI: 10.1148/radiol.2020191470.
[38]
WANG C, WU F, WANG F, et al. The association between tumor radiomic analysis and peritumor habitat-derived radiomic analysis on gadoxetate disodium-enhanced MRI with microvascular invasion in hepatocellular carcinoma[J]. J Magn Reson Imag, 2025, 61(3): 1428-1439. DOI: 10.1002/jmri.29523.
[39]
RENNE S L, SARCOGNATO S, SACCHI D, et al. Hepatocellular carcinoma: a clinical and pathological overview[J]. Pathologica, 2021, 113(3): 203-217. DOI: 10.32074/1591-951X-295.
[40]
WANG F, CHENG M, DU B B, et al. Predicting microvascular invasion in small (≤ 5cm) hepatocellular carcinomas using radiomics-based peritumoral analysis[J/OL]. Insights Imag, 2024, 15(1): 90 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/38530498/. DOI: 10.1186/s13244-024-01649-0.
[41]
ZHAO Y, ZHANG J, WANG N, et al. Intratumoral and peritumoral radiomics based on contrast-enhanced MRI for preoperatively predicting treatment response of transarterial chemoembolization in hepatocellular carcinoma[J/OL]. BMC Cancer, 2023, 23(1): 1026 [2025-02-15]. https://pubmed.ncbi.nlm.nih.gov/37875815/. DOI: 10.1186/s12885-023-11491-0.

上一篇 影像组学对乳腺癌腋窝淋巴结转移诊断价值的Meta分析
下一篇 联合临床指标和MR征象的列线图预测肝细胞癌肿瘤包绕型血管
  
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