磁共振影像组学分类乙肝人群的肝细胞癌和肝内肿块型胆管细胞癌

MRI based radiomics nomogram differentiate between hepatocellular carcinoma and intrahepatic mass-forming cholangiocarcinoma in HBV patients

摘要

目的 探讨基于多参数MRI影像组学列线图在乙肝患者中区分肝细胞癌(hepatocellular carcinoma, HCC)与肝内肿块型胆管细胞癌(intrahepatic mass-forming cholangiocarcinoma, IMCC)的效能。 方法 回顾性收集2016年8月至2022年9月于三家医院住院治疗的206例HCC和IMCC患者的MRI图像及临床资料,根据不同的就诊医院分成训练集(n=126)和外部测试集(n=80),分别提取三个序列(T1-FS、T2-FS、DWI)的影像组学特征,采用4种机器学习算法构建影像组学模型,挑选其中性能最优的影像组学模型用于后续分析。经过单因素和多因素分析筛选的临床特征作为独立预测因素,构建临床模型,并进一步联合影像组学模型建立临床-影像组学列线图。通过5折交叉验证进行超参数选择,并在外部测试集中进行评价。采用受试者工作特征(receiver operating characteristic, ROC)曲线评估模型诊断性能,德隆检验比较模型差异。 结果 经过ROC分析,基于线性支持向量机算法构建的影像组学模型在外部测试集表现最好,曲线下面积(area under curve, AUC)、灵敏度、特异性和准确性分别为0.929(95%CI:0. 872~0.986)、0.879、0.894和0.888;联合甲胎蛋白、糖类抗原199和性别的临床-影像组学列线图在测试集中的AUC、灵敏度、特异性和准确性分别为0.951(95%CI:0.910~1.000)、0.909、0.936和0.925,比较临床模型(AUC=0.822,95%CI:0.730~0.910, P=0.008)和影像组学模型(P=0.038)差异有统计学意义。 结论 影像组学列线图能够区分乙肝患者的IMCC与HCC,有助于指导治疗方法的选择,尤其是对于不能耐受增强扫描检查的患者。

Abstract

Objective To evaluate the performance of differentiating hepatocellular carcinoma(HCC)and intrahepatic mass-forming cholangiocarcinoma(IMCC)based on multiparametric MRI radiomics nomogram. Methods The retrospective, multicenter study involved 206 patients with either HCC or IMCC, who were enrolled from three hospitals between August 2016 and September 2022. Patients were divided into a training cohort(126 cases)and a testing cohort(80 cases)based on different hospitals. And the radiomics features of three sequences(T1-FS, T2-FS, and DWI)were extracted. The four machine learning algorithms were used to construct radiomics models. The optimal model was selected for further analysis. Independent clinical features were identified through univariate and multivariate analysis, which were then used to construct the clinical model. Subsequently, the radiomics nomogram was established by integrating independent clinical features with the selected radiomics model. The validation of all models was conducted through 5-fold cross-validation for hyperparameter optimization in the training cohort and evaluated in the external testing cohort. The receiver operating characteristic(ROC)curve was used to evaluate the model performance. The Delong test was used to compare the model differences. Results ROC analysis of the testing cohort demonstrated that the optimal radiomics model(T1+T2+DWI)by employing linear support vector machine achieved an area under curve(AUC)of 0.929(95%CI: 0.872-0.986), with sensitivity of 0.879, specificity of 0.894, and accuracy of 0.888. In conjunction with alpha fetoprotein, carbohydrate antigen 199 and sex, the clinical-radiomics nomogram exhibited superior performance, achieving an AUC of 0.951(95%CI: 0.910-1.000), with sensitivity of 0.909, specificity of 0.936, and accuracy of 0.925. These results were statistically superior to both the clinical model(AUC=0.822, 95%CI: 0.730-0.910; P=0.008)and the radiomics model(P=0.038), indicating significant added value from the integration of clinical and radiomic features. Conclusion The radiomics nomogram can distinguish between IMCC and HCC in patients with hepatitis B virus, thus facilitating more precise treatment plan, particularly for patients who are not able to tolerate enhanced scanning procedures.

关键词

影像组学 / 原发性肝癌 / 乙肝病毒 / 磁共振成像

Key words

Radiomics / Primary liver cancer / Hepatitis B virus / Magnetic resonance imaging

参考文献

[1] 何俊兴, 谢显文, 陈景艺. MRI影像特征对不典型肝细胞癌与肿块型肝内胆管癌的鉴别诊断价值[J]. 影像研究与医学应用, 2024, 8(23): 108-110.

[2] Banales JM, Rodrigues PM, Affò S, et al. Cholangiocarcinoma 2026: status quo, unmet needs and priorities[J]. Nat Rev Gastroenterol Hepatol, 2026, 23(1): 65-96.

[3] Choi SY, Kim YK, Min JH, et al. Added value of ancillary imaging features for differentiating scirrhous hepatocellular carcinoma from intrahepatic cholangiocarcinoma on gadoxetic acid-enhanced MR imaging[J]. Eur Radiol, 2018, 28(6): 2549-2560.

[4] Yoshimitsu K. Differentiation of two subtypes of intrahepatic cholangiocarcinoma: imaging approach[J]. Eur Radiol, 2019, 29(6): 3108-3110.

[5] 王洁. MRI延迟增强技术对肝内胆管细胞癌患者诊断特异性及敏感性分析[J]. 现代医用影像学, 2025, 34(5): 861-864.

[6] Rimola J, Forner A, Reig M, et al. Cholangiocarcinoma in cirrhosis: absence of contrast washout in delayed phases by magnetic resonance imaging avoids misdiagnosis of hepatocellular carcinoma[J]. Hepatology, 2009, 50(3): 791-798.

[7] 赖婳妤, 肖艳红, 温勇峰. 莫迪司核磁共振增强在早期肝癌患者诊断中的应用[J]. 影像研究与医学应用, 2020, 4(8): 209-211.

[8] 中华医学会影像技术分会国际交流学组. 肝胆特异性对比剂钆塞酸二钠增强MRI扫描方案专家共识[J]. 中华放射学杂志, 2019, 53(12): 1040-1044.

[9] International Communication Group,Imaging Technology Society of Chinese Medical Association. Expert consensus on hepatobiliary specific contrast agent gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid enhanced MRI scanning scheme[J]. Chinese Journal of Radiology, 2019, 53(12): 1040-1044.

[10] Si YQ, Wang XQ, Pan CC, et al. An efficient nomogram for discriminating intrahepatic cholangiocarcinoma from hepatocellular carcinoma: a retrospective study[J]. Front Oncol, 2022, 12: 833999. doi:10.3389/fonc.2022.833999

[11] Mahmoudi S, Bernatz S, Ackermann J, et al. Computed tomography radiomics to differentiate intrahepatic cholangiocarcinoma and hepatocellular carcinoma[J]. Clin Oncol, 2023, 35(5): 312-318.

[12] 吴思雨, 沈业隆, 王锡明. 影像组学预测原发性中枢神经系统淋巴瘤的Ki-67标记指数[J]. 山东大学学报(医学版), 2024, 62(11): 67-72.

[13] WU Siyu, SHEN Yelong, WANG Ximing. Radiomics predicts Ki-67 labeling index in primary central nervous system lymphomas[J]. Journal of Shandong University(Health Science), 2024, 62(11): 67-72.

[14] Zhang HP, Guo DJ, Liu H, et al. MRI-based radiomics models to discriminate hepatocellular carcinoma and non-hepatocellular carcinoma in LR-M according to LI-RADS version 2018[J]. Diagnostics, 2022, 12(5): 1043. doi:10.3390/diagnostics12051043

[15] 左立平, 蒋丰洋, 周斌彬, 等. 术前MRI在预测169例肝细胞肝癌微血管侵犯及早期复发的价值[J]. 山东大学学报(医学版), 2022, 60(3): 89-95.

[16] ZUO Liping, JIANG Fengyang, ZHOU Binbin, et al. Value of preoperative multiphase MRI for predicting microvascular invasion and early recurrence of 169 hepatocellular carcinoma[J]. Journal of Shandong University(Health Science), 2022, 60(3): 89-95.

[17] 靳新娟, 左立平, 邓展昊, 等. MRI影像组学对135例肝癌耐药蛋白PFKFB3的预测价值[J]. 山东大学学报(医学版), 2023, 61(6): 79-86.

[18] JIN Xinjuan, ZUO Liping, DENG Zhanhao, et al. Value of enhanced MRI radiomics in predicting the drug-resistant protein PFKFB3 in 135 cases of hepatocellular carcinoma[J]. Journal of Shandong University(Health Science), 2023, 61(6): 79-86.

[19] Wang SP, Wang XH, Yin XP, et al. Differentiating HCC from ICC and prediction of ICC grade based on MRI deep-radiomics: using lesions and their extended regions[J]. Phys Med, 2024, 120: 103322. doi:10.1016/j.ejmp.2024.103322

[20] Wu Q, Zhang T, Xu F, et al. MRI-based deep learning radiomics to differentiate dual-phenotype hepatocellular carcinoma from HCC and intrahepatic cholangiocarcinoma: a multicenter study[J]. Insights Imaging, 2025, 16(1): 27. doi:10.1186/s13244-025-01904-y

[21] 周灵玲, 郑丽云, 郭馨雨, 等. MRI影像组学分析技术在肝内胆管细胞癌诊断及预后评价中的研究进展[J]. 肝胆胰外科杂志, 2025, 37(2): 131-138.

[22] ZHOU Lingling, ZHENG Liyun, GUO Xinyu, et al. Advances in diagnosis and prognostic evaluation of intrahepatic cholangiocarcinoma based on MRI radiomics analysis techniques[J]. Journal of Hepatopancreatobi-liary Surgery, 2025, 37(2): 131-138.

[23] Peduzzi P, Concato J, Kemper E, et al. A simulation study of the number of events per variable in logistic regression analysis[J]. J Clin Epidemiol, 1996, 49(12): 1373-1379.

[24] Austin PC, Allignol A, Fine JP. The number of primary events per variable affects estimation of the subdistribution hazard competing risks model[J]. J Clin Epidemiol, 2017, 83: 75-84.

[25] van Smeden M, Moons KG, de Groot JA, et al. Sample size for binary logistic prediction models: beyond events per variable criteria[J]. Stat Methods Med Res, 2019, 28(8): 2455-2474.

[26] Li CQ, Huang H, Ruan SM, et al. An assessment of liver lesions using a combination of CEUS LI-RADS and AFP[J]. Abdom Radiol, 2022, 47(4): 1311-1320.

[27] 杨超豪, 马鹏飞, 梁志伟, 等. 基于血清学指标构建肝细胞癌与肝内胆管癌的鉴别诊断模型[J]. 河南外科学杂志, 2024, 30(4): 7-12.

[28] YANG Chaohao, MA Pengfei, LIANG Zhiwei, et al. Differential diagnosis model of hepatocellular carcinoma and intrahepatic cholangiocarcinoma based on serological indexes[J]. Henan Journal of Surgery, 2024, 30(4): 7-12.

[29] Tsuzaki J, Ueno A, Masugi Y, et al. Chronological changes in etiology, pathological and imaging findings in primary liver cancer from 2001 to 2020[J]. Jpn J Clin Oncol, 2025, 55(4): 362-371.

[30] Jiang K, Al-Diffhala S, Centeno BA. Primary liver cancers-part 1: histopathology, differential diagnoses, and risk stratification[J]. Cancer Control, 2018, 25(1): 1073274817744625.

[31] Liu N, Wu YK, Tao YY, et al. Differentiation of hepatocellular carcinoma from intrahepatic cholangiocarcinoma through MRI radiomics[J]. Cancers, 2023, 15(22): 5373. doi:10.3390/cancers15225373

[32] Bo ZY, Chen B, Yang Y, et al. Machine learning radiomics to predict the early recurrence of intrahepatic cholangiocarcinoma after curative resection: a multicentre cohort study[J]. Eur J Nucl Med Mol Imag, 2023, 50(8): 2501-2513.

[33] Mosconi C, Cucchetti A, Bruno A, et al. Radiomics of cholangiocarcinoma on pretreatment CT can identify patients who would best respond to radioembolisation[J]. Eur Radiol, 2020, 30(8): 4534-4544.

[34] Fiz F, Rossi N, Langella S, et al. Radiomic analysis of intrahepatic cholangiocarcinoma: non-invasive prediction of pathology data: a multicenter study to develop a clinical-radiomic model[J]. Cancers, 2023, 15(17): 4204. doi:10.3390/cancers15174204

[35] Shao S, Shan QG, Zheng N, et al. Role of intravoxel incoherent motion in discriminating hepatitis B virus-related intrahepatic mass-forming cholangiocarcinoma from hepatocellular carcinoma based on liver imaging reporting and data system v2018[J]. Cancer Biother Radiopharm, 2019, 34(8): 511-518.

[36] Sheng RF, Zhang YF, Wang HQ, et al. A multi-center diagnostic system for intrahepatic mass-forming cholangiocarcinoma based on preoperative MRI and clinical features[J]. Eur Radiol, 2024, 34(1): 548-559.

[37] Zhang XP, Jia NY, Wang YJ. Multi-input dense convolutional network for classification of hepatocellular carcinoma and intrahepatic cholangiocarcinoma[J]. Biomed Signal Process Control, 2023, 80: 104226. doi:10.1016/j.bspc.2022.104226

[38] Yang T, Yang Y, Hu SQ, et al. (18)F-FAPI-42 PET/CT for preoperatively identifying intrahepatic cholangiocarcinoma and hepatocellular carcinoma[J]. Quant Imaging Med Surg, 2025, 15(1): 741-751.

[39] Yeh H, Chiang CC, Yen TH. Hepatocellular carcinoma in patients with renal dysfunction: Pathophysiology, prognosis, and treatment challenges[J]. World J Gastroenterol, 2021, 27(26): 4104-4142.

[40] Bendszus M, Laghi A, Munuera J, et al. MRI gadolinium-based contrast media: meeting radiological, clinical, and environmental needs[J]. Magnetic Resonance Imaging, 2024: jmri. 29181. doi:10.1002/jmri.29181

[41] Shirata C, Hasegawa K, Kokudo T, et al. Liver resection for hepatocellular carcinoma in patients with renal dysfunction[J]. World J Surg, 2018, 42(12): 4054-4062.

[42] FAN R-E, CHANG K-W, HSIEH C-J, et al. LIBLINEAR: A Library for Large Linear Classification[J]. J Mach Learn Res, 2008, 9: 1871-1874.

[43] 解洪胜. 支持向量机在大规模数据分类中的应用[J]. 信息与电脑(理论版), 2017, 29(22): 44-45.

[44] XIE Hongsheng. Application of support vector machines in large-scale data classification[J]. China Computer & Communication, 2017, 29(22): 44-45.

[45] Hinselmann G, Rosenbaum L, Jahn A, et al. Large-scale learning of structure-activity relationships using a linear support vector machine and problem-specific metrics[J]. J Chem Inf Model, 2011, 51(2): 203-213.

基金资助

国家自然科学基金(82271993)

山东省自然科学基金(ZR2025QC1724)