[1]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:DOI:10.3969/j.issn.1005-202X.2026.02.015]
　HUANG Qian,ZHUANG Yinping,XU Peng,et al.Predicting microvascular invasion in hepatocellular carcinoma with multi-sequence MRI and a Swin Transformer-based deep learning model[J].Chinese Journal of Medical Physics,2026,43(2):245-254.[doi:DOI:10.3969/j.issn.1005-202X.2026.02.015]
点击复制

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()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
43卷
期数:
2026年第2期
页码:
245-254
栏目:
医学人工智能
出版日期:
2026-02-27

文章信息/Info

Title:
Predicting microvascular invasion in hepatocellular carcinoma with multi-sequence MRI and a Swin Transformer-based deep learning model
文章编号:
1005-202X（2026）02-0245-10
作者:
黄倩1; 庄银平1; 徐鹏2; 巩萍1
1.徐州医科大学医学影像学院， 江苏 徐州 221004； 2.徐州医科大学附属医院影像科， 江苏 徐州 221000
Author(s):
HUANG Qian1;  ZHUANG Yinping1;  XU Peng2;  GONG Ping1
1. School of Medical Imaging, Xuzhou Medical University, Xuzhou 221004, China 2. Department of Radiology, the Affiliated Hospital of Xuzhou Medical University, Xuzhou 221000, China
关键词:
肝细胞癌; 微血管侵犯; 多序列磁共振成像; Swin Transformer; 影像组学
Keywords:
Keywords: hepatocellular carcinoma microvascular invasion multi-sequence magnetic resonance imaging Swin Transformer radiomics
分类号:
R318；R735.7
DOI:
DOI:10.3969/j.issn.1005-202X.2026.02.015
文献标志码:
A
摘要:
目的：开发并验证基于多序列MRI的Swin Transformer（ST）深度学习模型在肝细胞癌（HCC）微血管侵犯（MVI）预测中的应用价值，为HCC患者的精准诊疗提供新的解决方案和客观、科学依据。方法：回顾性纳入徐州医科大学附属医院174例经手术病理确诊的HCC患者，收集患者的术前多序列MRI图像［动脉期（AP）、延迟期（DP）、脂肪抑制T2加权成像（T2WI-FS）、扩散加权成像（DWI）］，经过图像预处理和增强后，构建基于ST的DL模型，并将其与基于卷积神经网络（CNN）架构的DenseNet121、DenseNet169、ResNet34、ResNet50、VGG16、GoogleNet模型，基于Transformer架构的Vision Transformer（ViT）模型以及影像组学模型进行对比，计算模型的准确率、受试者工作特征曲线下面积（AUC）、敏感度、特异度，通过混淆矩阵和决策曲线分析（DCA）评估性能。为提高模型的可解释，采用梯度加权类激活映射（Grad-CAM）可视化模型关注区域。结果：ST模型在4个序列的MVI预测任务中总体性能最好，特别是DP序列上预测结果最优，其准确度、AUC（95%CI）、敏感度和特异度分别为0.944、0.993（0.985~0.998）、0.984和0.904，显著优于其他深度学习模型和影像组学模型，DCA图进一步证实其卓越的临床应用价值。Grad-CAM结果显示ST模型能够有效关注肿瘤相关区域。结论：ST在不同序列MRI图像上的各项指标均表现出优异的预测性能，可为临床诊断、治疗决策及预后评估提供有力的辅助工具。
Abstract:
Abstract: Objective To develop and validate a Swin Transformer (ST)-based deep learning (DL) model using multi-sequence magnetic resonance imaging (MRI) for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC), thereby providing a novel solution and objective scientific evidence to facilitate the precision diagnosis and treatment for HCC patients. Methods A retrospective analysis was performed on 174 patients with surgically and pathologically confirmed HCC who were admitted to the Affiliated Hospital of Xuzhou Medical University. Preoperative multi-sequence MRI images, including arterial phase, delayed phase (DP), fat-suppressed T2-weighted imaging, and diffusion-weighted imaging, were collected for each patient. After image preprocessing and enhancement, an ST-based DL model was established and then compared with models based on convolutional neural network architectures, including DenseNet121, DenseNet169, ResNet34, ResNet50, VGG16, and GoogleNet, as well as a Transformer-based Vision Transformer model and a radiomics model. Model accuracy, area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were calculated, and the model performance was evaluated using confusion matrices and decision curve analysis. To enhance model interpretability, gradient-weighted class activation mapping (Grad-CAM) was employed to visualize the regions of interest focused on by the models. Results The ST model demonstrated the optimal overall performance among the 4 sequences in MVI prediction, with particularly outstanding results on the DP sequence. Specifically, it achieved an accuracy of 0.944, an AUC of 0.993 (95% CI: 0.985-0.998), a sensitivity of 0.984, and a specificity of 0.904, significantly outperforming other DL models and the radiomics model. Decision curve analysis further confirmed its superior potential for clinical application. Grad-CAM visualization indicated that the ST model effectively concentrated on tumor-associated regions. Conclusion The ST model exhibits excellent predictive performance across multiple MRI sequences and can served as a powerful auxiliary tool to support clinical diagnosis, treatment decision-making, and prognostic evaluation.

相似文献/References:

[1]韩柱君,巩贯忠,路玉昆,等. 原发性肝细胞癌的CT影像组学分析[J].中国医学物理学杂志,2018,35(12):1426.[doi:DOI:10.3969/j.issn.1005-202X.2018.12.011]
　HAN Zhujun,GONG Guanzhong,LU Yukun,et al. CT radiomic features of primary hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2018,35(2):1426.[doi:DOI:10.3969/j.issn.1005-202X.2018.12.011]
[2]闫冰,吴爱东,张洪波,等. 肝细胞癌共面和非共面容积旋转调强放疗与螺旋断层放疗的剂量学研究[J].中国医学物理学杂志,2019,36(8):877.[doi:DOI:10.3969/j.issn.1005-202X.2019.08.003]
　YAN Bing,WU Aidong,ZHANG Hongbo,et al. Dosimetric study on coplanar and noncoplanar volumetric modulated arc therapy and helical tomotherapy for hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2019,36(2):877.[doi:DOI:10.3969/j.issn.1005-202X.2019.08.003]
[3]刘海峰,许永生,刘钊,等.钆塞酸二钠增强磁共振和弥散加权成像诊断肝细胞癌TACE术后存活或复发病灶的价值[J].中国医学物理学杂志,2020,37(5):561.[doi:10.3969/j.issn.1005-202X.2020.05.006]
　LIU Haifeng,XU Yongsheng,LIU Zhao,et al.Diagnostic value of Gd-EOB-DTPA enhanced MRI and DWI in residual or recurrent lesionsafter TACE for hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2020,37(2):561.[doi:10.3969/j.issn.1005-202X.2020.05.006]
[4]吴林耿,陈猛,苟庆,等.中性粒细胞/淋巴细胞比值对晚期肝细胞癌FOLFOX肝动脉灌注化疗近期疗效的影响[J].中国医学物理学杂志,2021,38(5):602.[doi:DOI:10.3969/j.issn.1005-202X.2021.05.015]
　WU Lingeng,CHEN Meng,et al.Effects of neutrophil to lymphocyte ratio on short-term efficacy of FOLFOX hepatic arterial infusion chemotherapy for advanced hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2021,38(2):602.[doi:DOI:10.3969/j.issn.1005-202X.2021.05.015]
[5]魏佳赟,姚佳,李汛.细胞外基质刚度在肝细胞癌中的研究进展[J].中国医学物理学杂志,2021,38(6):770.[doi:DOI:10.3969/j.issn.1005-202X.2021.06.020]
　WEI Jiayun,YAO Jia,et al.Advances of research on extracellular matrix stiffness in hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2021,38(2):770.[doi:DOI:10.3969/j.issn.1005-202X.2021.06.020]
[6]范义湘,易紫薇,胡煜麟,等.NIS基因转染对肝细胞癌NIS蛋白表达及摄碘功能的影响[J].中国医学物理学杂志,2021,38(10):1219.[doi:DOI:10.3969/j.issn.1005-202X.2021.10.007]
　FAN Yixiang,YI Ziwei,HU Yulin,et al.Effects of NIS gene transfection on NIS protein expression and iodine uptake in hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2021,38(2):1219.[doi:DOI:10.3969/j.issn.1005-202X.2021.10.007]
[7]谢孟臻,吴斌,许立强,等.双层探测器光谱CT虚拟平扫中重建层厚对肝细胞癌显示的影响[J].中国医学物理学杂志,2022,39(12):1495.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.007]
　XIE Mengzhen,WU Bin,XU Liqiang,et al.Effect of reconstructed slice thickness on the display of hepatocellular carcinoma in virtual non-contrast scan of dual-layer detector spectral CT[J].Chinese Journal of Medical Physics,2022,39(2):1495.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.007]
[8]李申超,沈树成,赵汝彬,等.肝细胞癌常规与缩小射束夹角放疗计划的肝段剂量学比较[J].中国医学物理学杂志,2025,42(12):1556.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.003]
　LI Shenchao,SHEN Shucheng,et al.Dosimetric comparison of hepatic segments between conventional and beam angle-reduced radiotherapy plans for hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2025,42(2):1556.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.003]
[9]刘杨军,刘淑珍,李鹏,等.基于MSCT特征构建肝细胞癌根治性切除术后复发的预测模型[J].中国医学物理学杂志,2026,43(2):261.[doi:DOI:10.3969/j.issn.1005-202X.2026.02.017]
　LIU Yangjun,LIU Shuzhen,LI Peng,et al.Construction of a predictive model for hepatocellular carcinoma recurrence after radical resection based on MSCT features[J].Chinese Journal of Medical Physics,2026,43(2):261.[doi:DOI:10.3969/j.issn.1005-202X.2026.02.017]
[10]陈昭,张宇,周乐,等.基于钆赛酸二钠增强MRI的影像组学与深度迁移学习预测肝细胞癌术前微血管侵犯[J].中国医学物理学杂志,2025,42(10):1353.[doi:DOI:10.3969/j.issn.1005-202X.2025.10.013]
　CHEN Zhao,ZHANG Yu,ZHOU Le,et al.Radiomics and deep transfer learning based on gadoxetic acid disodium-enhanced MRI for predicting preoperative microvascular invasion in hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2025,42(2):1353.[doi:DOI:10.3969/j.issn.1005-202X.2025.10.013]

备注/Memo

备注/Memo:
【收稿日期】2025-10-21 【基金项目】国家自然科学基金（82001987）；江苏省研究生科研与实践创新计划（KYCX25_3266） 【作者简介】黄倩，硕士研究生，研究方向：肝癌人工智能分析，E-mail: 303103110587@stu.xzhmu.edu.cn 【通信作者】巩萍，博士，副教授，硕士生导师，研究方向：医学图像处理与人工智能分析，E-mail: gongping@xzhmu.edu.cn
更新日期/Last Update: 2026-01-27