[1]傅力嘉,林艳萍,李娜,等.基于跨模态生成与交叉注意力的乳腺超声诊断[J].中国医学物理学杂志,2026,43(7):889-897.[doi:DOI:10.3969/j.issn.1005-202X.2026.07.008]
　FU Lijia,LIN Yanping,LI Na,et al.Breast ultrasound diagnosis based on cross-modal generation and cross-attention[J].Chinese Journal of Medical Physics,2026,43(7):889-897.[doi:DOI:10.3969/j.issn.1005-202X.2026.07.008]
点击复制

基于跨模态生成与交叉注意力的乳腺超声诊断()

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

卷:
43卷
期数:
2026年第7期
页码:
889-897
栏目:
医学影像物理
出版日期:
2026-07-22

文章信息/Info

Title:
Breast ultrasound diagnosis based on cross-modal generation and cross-attention
文章编号:
1005-202X（2026）07-0889-09
作者:
傅力嘉1; 林艳萍1; 李娜2; 贾超3; 4; 李刚3; 4; 杜联芳3; 4; 李凡4; 5
1.上海交通大学机械与动力工程学院， 上海 200240；2.上海理工大学健康科学与工程学院， 上海 200093；3.上海市第一人民医院超声医学科， 上海 201620；4.上海交通大学医学院， 上海 200025；5.上海市胸科医院超声科， 上海 200030
Author(s):
FU Lijia1;  LIN Yanping1;  LI Na2;  JIA Chao3;  4;  LI Gang3;  4;  DU Lianfang3;  4;  LI Fan4;  5
1. School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China 2. School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China 3. Department of Ultrasound, Shanghai General Hospital, Shanghai 201620, China 4. School of Medicine, Shanghai Jiao Tong University, Shanghai 200025, China 5. Department of Ultrasound, Shanghai Chest Hospital, Shanghai 200030, China
关键词:
乳腺癌; 乳腺超声; 超声造影; 辅助诊断; 生成对抗网络; 跨模态生成; 交叉注意力; 特征融合
Keywords:
Keywords: breast cancer breast ultrasound contrast-enhanced ultrasound auxiliary diagnosis generative adversarial network cross-modal generation cross-attention feature fusion
分类号:
R318；TP391.41
DOI:
DOI:10.3969/j.issn.1005-202X.2026.07.008
文献标志码:
A
摘要:
的：针对常规B超在乳腺癌筛查中因读图者间一致性不佳、微小病灶恶性征象不典型导致的准确性受限问题，以及超声造影因成本高、操作复杂而难以普及的临床矛盾，提出一种“分割-生成-分类”三阶段级联诊断框架，旨在仅利用常规B超图像实现高精度的乳腺肿瘤良恶性诊断。方法：首先，采用CNN-Transformer架构的分割网络从B超图像中提取肿瘤区域。接着，构建基于条件生成对抗网络的跨模态生成模型，利用预训练的ResNet-34编码器从分割后的B超中生成对应的超声造影图像。最后，设计以Inception-v3为骨干的双分支交叉注意力网络，通过动态特征融合实现对真实B超与生成造影信息的深度交互，完成对肿瘤良恶性的精确诊断。结果：在包含677例患者的乳腺超声数据集上进行验证，该框架的受试者工作特征曲线下面积（AUC）达到0.930，显著优于多个单模态B超基线模型（最佳AUC为0.868）。消融实验证实病灶分割、跨模态生成以及交叉注意力特征融合模块对提升诊断性能的必要贡献。结论：本研究提出的框架验证了通过深度学习方法生成并利用多模态超声信息进行高精度诊断的可行性与有效性，能够以低成本、高效率的方式提供多模态诊断增益，有望为乳腺癌临床筛查提供有力的辅助决策支持，降低乳腺癌临床筛查的漏诊与误诊风险。
Abstract:
Abstract: Objective To address the limited accuracy of conventional B-mode ultrasound (US) in breast cancer screening, resulting from poor inter-reader agreement and atypical malignant signs of tiny lesions, and resolve the clinical dilemma where contrast-enhanced ultrasound (CEUS) is hard to popularize due to its high cost and operational complexity, this study proposes a three-stage cascaded diagnostic framework of "segmentation-generation-classification" to achieve high-precision diagnosis of benign and malignant breast tumors relying solely on conventional US images. Methods A CNN-Transformer framework-based segmentation network was first employed to extract tumor regions from US images. Subsequently, a cross-modal generation model based on a conditional generative adversarial network was constructed, utilizing a pre-trained ResNet-34 encoder to synthesize corresponding CEUS images from the segmented US images. Finally, a dual-branch cross-attention network with Inception-v3 as the backbone was designed to achieve deep interaction between the real US and the generated CEUS information through dynamic feature fusion, thereby accomplishing accurate benign and malignant tumor classification. Results Validated on a breast ultrasound dataset including 677 patients, the proposed framework achieved an area under the receiver operating characteristic curve (AUC) of 0.930, significantly outperforming several single-modal US baseline models which reached a maximum AUC of 0.868. Ablation experiments verified the essential contributions of lesion segmentation, cross-modal generation, and cross-attention feature fusion modules in enhancing diagnostic performance. Conclusion This study validates the feasibility and effectiveness of generating and utilizing multimodal ultrasound information with deep learning for high-precision diagnosis. The proposed framework provides multimodal diagnostic benefits in a cost-effective and highly efficient way, holding great promise to provide robust auxiliary decision support for clinical breast cancer screening and reduce the risks of missed diagnoses and misdiagnoses.

相似文献/References:

[1]柏 晗,王 丽,刘旭红,等.左乳腺5种不同主野分布IMRT计划的剂量比较分析[J].中国医学物理学杂志,2015,32(02):210.[doi:10.3969/j.issn.1005-202X.2015.02.013]
[2]张晋建,黎 静,文 婷,等.Arc CHECK在乳腺癌螺旋断层放疗计划剂量验证中的应用[J].中国医学物理学杂志,2015,32(02):244.[doi:10.3969/j.issn.1005-202X.2015.02.021]
[3]白 雪,杜向慧,王升晔,等.半野技术结合非共面照射技术在乳腺癌胸锁联合照射中的应用[J].中国医学物理学杂志,2014,31(03):4849.[doi:10.3969/j.issn.1005-202X.2014.03.001]
[4]赵永亮,储开岳,吴建亭,等.两种不同调强技术配合主动呼吸控制系统在乳腺癌放疗中的比较[J].中国医学物理学杂志,2014,31(03):4862.[doi:10.3969/j.issn.1005-202X.2014.03.004]
[5]花威,张西志,张先稳,等.RapidArc在乳腺癌术后放疗中的应用[J].中国医学物理学杂志,2013,30(02):3984.[doi:10.3969/j.issn.1005-202X.2013.02.005]
[6]徐金济,成俊,李伟,等.左侧乳腺癌改良根治术后逆向调强放射治疗的摆位误差分析[J].中国医学物理学杂志,2013,30(05):4358.[doi:10.3969/j.issn.1005-202X.2013.05.004]
[7]孟慧鹏,孙小喆,孙劲松,等.右乳腺癌术后放疗中VMAT和dIMRT的剂量学比较[J].中国医学物理学杂志,2013,30(05):4387.[doi:10.3969/j.issn.1005-202X.2013.05.011]
[8]刘旭红,陈晓,艾毅钦,等.早期左乳腺癌保乳术后仰卧位和俯卧位照射剂量学比较与分析[J].中国医学物理学杂志,2015,32(05):737.[doi:doi:10.3969/j.issn.1005-202X.2015.05.028]
　[J].Chinese Journal of Medical Physics,2015,32(7):737.[doi:doi:10.3969/j.issn.1005-202X.2015.05.028]
[9]张艺宝,吴昊,李莎,等.临床前验证与几何对比分析基于图谱库的危及器官自动勾画[J].中国医学物理学杂志,2015,32(06):761.[doi:doi:10.3969/j.issn.1005-202X.2015.06.001]
　[J].Chinese Journal of Medical Physics,2015,32(7):761.[doi:doi:10.3969/j.issn.1005-202X.2015.06.001]
[10]段云龙,杨涛,曲宝林,等.乳腺癌保乳切线野三维适形、切线野正向调强及逆向调强放疗的剂量学评估[J].中国医学物理学杂志,2015,32(06):841.[doi:doi:10.3969/j.issn.1005-202X.2015.06.017]
　[J].Chinese Journal of Medical Physics,2015,32(7):841.[doi:doi:10.3969/j.issn.1005-202X.2015.06.017]
[11]代妮娜,张文君.基于多模态融合技术提升乳腺超声与磁共振成像联合诊断乳腺癌的效能[J].中国医学物理学杂志,2026,43(3):317.[doi:DOI:10.3969/j.issn.1005-202X.2026.03.006]
　DAI Nina,ZHANG Wenjun.Improving the diagnostic efficacy of breast cancer by integrating breast ultrasound and magnetic resonance imaging with multimodal fusion technology[J].Chinese Journal of Medical Physics,2026,43(7):317.[doi:DOI:10.3969/j.issn.1005-202X.2026.03.006]

备注/Memo

备注/Memo:
【收稿日期】2025-11-25 【基金项目】国家自然科学基金（52175020）；上海市自然科学基金（21ZR1451400） 【作者简介】傅力嘉，硕士研究生，研究方向：医学图像处理，E-mail: fulili@sjtu.edu.cn 【通信作者】李凡，博士，主任医师，副研究员，博士生导师，研究方向：超声生物治疗、超声造影及多模态超声诊疗，E-mail: medicineli@sjtu.edu.cn
更新日期/Last Update: 2026-07-22