[1]李好,杨智慧,李丰森.基于注意力增强与边缘感知的脑肿瘤MRI跨模态生成方法[J].中国医学物理学杂志,2026,43(1):65-75.[doi:DOI:10.3969/j.issn.1005-202X.2026.01.009]
　LI Hao,YANG Zhihui,et al.Cross-modality brain tumor MRI image synthesis method based on attention enhancement and edge awareness[J].Chinese Journal of Medical Physics,2026,43(1):65-75.[doi:DOI:10.3969/j.issn.1005-202X.2026.01.009]
点击复制

基于注意力增强与边缘感知的脑肿瘤MRI跨模态生成方法()

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

卷:
43卷
期数:
2026年第1期
页码:
65-75
栏目:
医学影像物理
出版日期:
2026-01-26

文章信息/Info

Title:
Cross-modality brain tumor MRI image synthesis method based on attention enhancement and edge awareness
文章编号:
1005-202X（2026）01-0065-11
作者:
李好1; 2; 杨智慧1; 2; 李丰森1; 2
1.甘肃中医药大学医学信息工程学院， 甘肃 兰州 730000； 2.甘肃省人民医院/甘肃中医药大学第一临床医学院， 甘肃 兰州 730000
Author(s):
LI Hao1;  2;  YANG Zhihui1;  2;  LI Fengsen1;  2
1. School of Medical Information Science and Engineering, Gansu University of Chinese Medicine, Lanzhou 730000, China 2. Gansu Provincial Hospital/First School of Clinical Medicine, Gansu University of Chinese Medicine, Lanzhou 730000, China
关键词:
生成对抗网络; 脑肿瘤MRI图像生成; 注意力机制; 边缘感知; 梯度正则化
Keywords:
generative adversarial network brain tumor MRI image synthesis attention mechanism edge awareness gradient regularization
分类号:
R318
DOI:
DOI:10.3969/j.issn.1005-202X.2026.01.009
文献标志码:
A
摘要:
目的：规避脑肿瘤MRI成像过程中存在的时间成本高、伪影多和模态获取不全等问题，研究一种高质量的跨模态脑肿瘤MRI图像生成方法。方法：提出一种融合注意力机制与边缘感知的配准生成对抗网络（AE-RegGAN），对T1模态到T2模态图像的跨模态合成，在生成器中引入CoordAttention模块以增强关键区域感知，并结合Sobel边缘检测以强化肿瘤边界表达；在判别器中加入梯度惩罚正则化以提升训练稳定性并缓解模式崩溃问题。结果：在对5 760例脑肿瘤MRI数据训练、768例测试中，AE-RegGAN相较于原始RegGAN在局部肿瘤区域的峰值信噪比（PSNR）提升0.51 dB，结构相似性指数（SSIM）提升0.029；在全局图像上PSNR提升0.900 dB，SSIM提升0.032。全局图像配对t检验结果显示平均绝对误差（P=0.026 4）、PSNR（P<0.000 1）、SSIM（P<0.000 1）指标差异均有统计学意义。消融实验进一步验证了注意力与边缘感知模块的有效性。结论：AE-RegGAN在多模态脑部MRI图像合成中表现出更优的结构保持能力与病灶敏感性，为辅助诊断提供了稳定、可信的图像补全方案。
Abstract:
Objective To develop a high-quality cross-modality synthesis method for magnetic resonance imaging (MRI) of brain tumor, tackling the challenges of high acquisition time cost, multiple artifacts, and incomplete modality availability in MRI. Methods A registration-enhanced generative adversarial network named AE?egGAN integrating attention mechanism and edge awareness was developed for cross-modality synthesis from T1-weighted to T2-weighted MRI images. In the generator, a CoordAttention module was embedded to enhance the perception of key regions, while a Sobel-based edge detection mechanism was incorporated to strengthen the representation of tumor boundaries. In the discriminator, gradient penalty regularization was adopted to improve training stability and mitigate mode collapse. Results The model was trained on 5 760 brain tumor MRI samples and evaluated on 768 test cases. AE?egGAN outperformed the original RegGAN, exhibiting an improvement of 0.51 dB in peak signal-to-noise ratio (PSNR) and 0.029 in structural similarity index measure (SSIM) in local tumor regions. For global image, the PSNR and SSIM were enhanced by 0.900 dB and 0.032, respectively. Paired t?est results for global image demonstrated statistically significant improvements in mean absolute error (P=0.026 4), PSNR (P<0.000 1), and SSIM (P<0.000 1). Ablation experiments further validated the effectiveness of the attention and edge awareness modules. Conclusion AE?egGAN exhibits superior structural preservation capability and lesion sensitivity in multimodal brain MRI synthesis, offering a stable and reliable image completion solution to support clinical diagnosis.

相似文献/References:

[1]弥佳,周宇佳,冯前进.基于生成对抗网络的无监督动态对比增强磁共振分解方法[J].中国医学物理学杂志,2021,38(2):186.[doi:DOI:10.3969/j.issn.1005-202X.2021.02.011]
　MI Jia,ZHOU Yujia,FENG Qianjin.Unsupervised image decomposition of DCE-MRI based on generative adversarial networks[J].Chinese Journal of Medical Physics,2021,38(1):186.[doi:DOI:10.3969/j.issn.1005-202X.2021.02.011]
[2]郭翌,吴香奕,吴茜,等.基于循环一致生成对抗网络的多模态影像刚性配准[J].中国医学物理学杂志,2021,38(2):198.[doi:DOI:10.3969/j.issn.1005-202X.2021.02.013]
　GUO Yi,WU Xiangyi,WU Qian,et al.Rigid registration of multimodal images based on CycleGAN[J].Chinese Journal of Medical Physics,2021,38(1):198.[doi:DOI:10.3969/j.issn.1005-202X.2021.02.013]
[3]侯振峰,耿长冉,田锋,等.基于蒙特卡罗方法与人体模型的硼中子俘获治疗硼浓度在线监测方法[J].中国医学物理学杂志,2023,40(3):278.[doi:DOI:10.3969/j.issn.1005-202X.2023.03.004]
　HOU Zhenfeng,GENG Changran,,et al.Online BNCT boron concentration monitoring method based on Monte Carlo simulation and human body model[J].Chinese Journal of Medical Physics,2023,40(1):278.[doi:DOI:10.3969/j.issn.1005-202X.2023.03.004]
[4]徐盼盼,张东,袁达龙.基于生成对抗网络的多模态MR图像缺失模态合成[J].中国医学物理学杂志,2023,40(7):827.[doi:DOI:10.3969/j.issn.1005-202X.2023.07.006]
　XU Panpan,ZHANG Dong,YUAN Dalong.Incomplete multimodality MR image synthesis based on generative adversarial network[J].Chinese Journal of Medical Physics,2023,40(1):827.[doi:DOI:10.3969/j.issn.1005-202X.2023.07.006]
[5]周金治,胡震,郭莉莉,等.基于GAN-DAUnet的肝脏CT图像肿瘤分割算法[J].中国医学物理学杂志,2023,40(8):971.[doi:DOI:10.3969/j.issn.1005-202X.2023.08.008]
　ZHOU Jinzhi,HU Zhen,et al.Liver CT image tumor segmentation algorithm based on GAN-DAUnet[J].Chinese Journal of Medical Physics,2023,40(1):971.[doi:DOI:10.3969/j.issn.1005-202X.2023.08.008]
[6]吴书裕,王伟,阳露,等.基于自监督生成对抗网络的锥形束CT合成伪CT研究[J].中国医学物理学杂志,2023,40(11):1356.[doi:DOI:10.3969/j.issn.1005-202X.2023.11.007]
　WU Shuyu,WANG Wei,YANG Lu,et al.Synthetic CT generation from CBCT images using self-attention generative adversarial network[J].Chinese Journal of Medical Physics,2023,40(1):1356.[doi:DOI:10.3969/j.issn.1005-202X.2023.11.007]
[7]尹智贤,夏克文,张昭,等.融合上下文注意力的两段式生成对抗网络的肺结节图像生成与分类[J].中国医学物理学杂志,2024,41(12):1517.[doi:DOI:10.3969/j.issn.1005-202X.2024.12.009]
　YIN Zhixian,XIA Kewen,et al.Synthesis and classification of pulmonary nodules using two-stage-based generative adversarial network incorporating contextual transformer[J].Chinese Journal of Medical Physics,2024,41(1):1517.[doi:DOI:10.3969/j.issn.1005-202X.2024.12.009]
[8]李芳华,丁寿亮,李永宝,等.基于Transformer的生成对抗网络用于鼻咽癌MRI生成伪CT[J].中国医学物理学杂志,2025,42(6):701.[doi:DOI:10.3969/j.issn.1005-202X.2025.06.001]
　LI Fanghua,DING Shouliang,LI Yongbao,et al.Synthetic CT generation from NPC MRI using Transformer-based generative adversarial network[J].Chinese Journal of Medical Physics,2025,42(1):701.[doi:DOI:10.3969/j.issn.1005-202X.2025.06.001]
[9]林洁,李彩霞,黎海霞,等.运用循环一致生成对抗网络对肥胖肺结核患者超低剂量CT图像进行降噪[J].中国医学物理学杂志,2025,42(12):1601.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.009]
　LIN Jie,LI Caixia,LI Haixia,et al.Denoising ultra-low-dose CT images of obese patients with pulmonary tuberculosis using cycle-consistent generative adversarial network[J].Chinese Journal of Medical Physics,2025,42(1):1601.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.009]
[10]邢雄飞,樊星佳,李睿.基于生成对抗网络的细胞图像超分辨重建[J].中国医学物理学杂志,2026,43(1):52.[doi:DOI:10.3969/j.issn.1005-202X.2026.01.007]
　XING Xiongfei,FAN Xingjia,LI Rui.Super-resolution reconstruction of cell images based on generative adversarial network[J].Chinese Journal of Medical Physics,2026,43(1):52.[doi:DOI:10.3969/j.issn.1005-202X.2026.01.007]

备注/Memo

备注/Memo:
【收稿日期】2025-05-12 【基金项目】甘肃省科技计划项目（25JRZA204） 【作者简介】李好，硕士研究生，研究方向：医学图像生成，E-mail: 1823859315@qq.com 【通信作者】李丰森，硕士生导师，正高级工程师，研究方向：医学信息工程，E-mail: 1641673807@qq.com
更新日期/Last Update: 2026-01-27