[1]彭晨阳,何立风,王梦溪,等.SDA U-Mamba: 基于频域动态特征融合与双极路由注意力的医学图像分割[J].智能系统学报,2026,21(1):284-294.[doi:10.11992/tis.202508032]
PENG Chenyang,HE Lifeng,WANG Mengxi,et al.SDA U-Mamba: spectral-domain dynamic fusion and bipolar routing attention for medical image segmentation[J].CAAI Transactions on Intelligent Systems,2026,21(1):284-294.[doi:10.11992/tis.202508032]
点击复制
PENG Chenyang,HE Lifeng,WANG Mengxi,et al.SDA U-Mamba: spectral-domain dynamic fusion and bipolar routing attention for medical image segmentation[J].CAAI Transactions on Intelligent Systems,2026,21(1):284-294.[doi:10.11992/tis.202508032]
SDA U-Mamba: 基于频域动态特征融合与双极路由注意力的医学图像分割
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
21
期数:
2026年第1期
页码:
284-294
栏目:
人工智能院长论坛
出版日期:
2026-03-05
- Title:
- SDA U-Mamba: spectral-domain dynamic fusion and bipolar routing attention for medical image segmentation
- Keywords:
- Mamba; medical image segmentation; state space model; spectral-dynamic feature fusion; Fourier transform; attention mechanism; U-Net architecture; multi-scale feature modeling.
- 分类号:
- TP391.4
- 摘要:
- 以Mamba为代表的状态空间模型(state space models，SSM)凭借其出色的长程依赖建模能力与较低的线性计算复杂度，在医学图像分割领域展现出广阔的应用前景。但该类方法对图像逐像素展平的处理方式会破坏图像空间结构，导致局部细节提取不足，且缺乏对器官与病灶的聚焦机制，在复杂背景下易引入冗余背景信息。为解决上述问题，本文提出了一种融合谱域动态特征与注意力机制的医学图像分割网络——频域动态注意力U型Mamba(spectral dynamic attention U-Mamba，SDA U-Mamba)。该网络采用分层U型结构设计，分别从空间连续性建模和区域聚焦能力两方面对Mamba进行优化。在网络浅层设计Mamba空频注意力模块，通过融合空域卷积、频域变换和金字塔自注意力结构以提升模型的局部信息感知能力与多尺度上下文建模效果；在网络深层引入双极路由注意力模块，通过动态路由选择与稀疏激活机制增强模型对医学图像器官或病灶的表征。实验结果表明，SDA U-Mamba在BUSI、CVC-ClinicDB与CHAOS-Liver这3个公开医学数据集上的分割性能显著优于当前主流方法，平均交并比(intersection over union，IoU)提升2.61%。本文所提算法可用于临床医学图像分割。
- Abstract:
- State space models(SSM) represented by Mamba have demonstrated broad application prospects in medical image segmentation owing to their excellent long-range dependency modeling and low linear computational complexity. However, the pixel-wise flattening operation used in these methods destroys the spatial structure of images, leading to insufficient extraction of local details. Additionally, they lack a mechanism to focus on organs and lesions, making them prone to introducing redundant background information in medical images with complex contexts. To address these issues,this study proposes spectral dynamic attention U-Mamba(SDA U-Mamba), a medical image segmentation network that integrates spectral-domain dynamic features with attention mechanisms. The network adopts a hierarchical U-shaped architecture, optimizing Mamba in two aspects: spatial continuity modeling and regional focusing capability. Specifically, a Mamba Spatial-Frequency Attention module is introduced into the network’s shallow layers. This module enhances the model’s local information perception and multi-scale context modeling performance by fusing spatial convolution, frequency-domain transformation, and a pyramid self-attention structure. Furthermore, a Bipolar Routing Attention module is introduced in the deep layers, strengthening the network’s representation of organs and lesions in medical images through dynamic routing selection and sparse activation mechanisms. Experimental results show that SDA U-Mamba achieved significantly better segmentation performance than current state-of-the-art methods across three public medical datasets (BUSI, CVC-ClinicDB, and CHAOS-Liver), with an average improvement in intersection over union(IoU) of 2.61%. The results indicate that the proposed algorithm can be applied to clinical medical image segmentation.
- 参考文献/References:
-
[1] 周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017(6): 1229-1251 ZHOU Feiyan, JIN Linpeng, DONG Jun. Review of convolutional neural network[J]. Chinese journal of computers, 2017(6): 1229-1251
[2] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. Pattern recognition letters, 2020, 133: 157-165
[3] HUANG Gao, LIU Zhuang, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2261-2269.
[4] TAN Mingxing, LE Q V. Efficientnet: rethinking model scaling for convolutional neural networks[C]//International Conference on Machine Learning. Long Beach: PMLR, 2019.
[5] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[C]//International Conference on Le-arning Representations. Virtual: ICLR, 2021.
[6] LEI Tao, SUN Rui, DU Xiaogang, et al. SGU-Net: shape-guided ultralight network for abdominal image segmentation[J]. IEEE journal of biomedical and health informatics, 2023, 27(3): 1431-1442
[7] 周丽娟, 毛嘉宁. 视觉Transformer识别任务研究综述[J]. 中国图象图形学报, 2023, 28(10): 2969-3003 ZHOU Lijuan, MAO Jianing. Vision Transformer-based recognition tasks: a critical review[J]. Journal of image and graphics, 2023, 28(10): 2969-3003
[8] GU A, GOEL K, R? C. Efficiently modeling long sequences with structured state spaces[C]//International Conference on Learning Representations. Virtual: ICLR, 2022.
[9] GU A, DAO T. Mamba: linear-time sequence modeling with selective state spaces[EB/OL]. (2023-12-01)[2025-04-20]. https://arxiv.org/abs/2312.00752.
[10] JIAO Jianbin, LIU Yue, LIU Yunfan, et al. VMamba: visual state space model[C]//Advances in Neural Information Processing Systems. Vancouver: Neural Information Processing Systems Foundation, Inc. , 2024.
[11] ZHU Lianghui, LIAO Bencheng, ZHANG Qian, et al. Vision mamba: efficient visual representation learning with bidirectional state space model[C]//Proceedings of the 41st International Conference on Machine Learning. Baltimore: PMLR, 2024.
[12] HE Xuanhua, CAO Ke, ZHANG Jie, et al. Pan-Mamba: effective pan-sharpening with state space model[J]. Information fusion, 2025, 115: 102779
[13] PEI Xiaohuan, HUANG Tao, XU Chang. Efficientvmamba: atrous selective scan for light weight visual mamba[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver: AAAI, 2025.
[14] CHENG Zihan, GUO Jintao, ZHANG Jian, et al. Mamba-sea: a mamba-based framework with global-to-local sequence augmentation for generalizable medical image segmentation[J]. IEEE transactions on medical imaging, 2025, 44(9): 3741-3755
[15] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer Assisted Intervention. Munich: MICCAI Society, 2015.
[16] ZHOU Zongwei, RAHMAN SIDDIQUEE M M, TAJBAKHSH N, et al. UNet++: a nested UNet architecture for medical image segmentation[C]//Proceedings of the 4th International Workshop on Deep Learning in Medical Image Analysis. Granada: DLMIA, 2018.
[17] HUANG Huimin, LIN Lanfen, TONG Ruofeng, et al. UNet 3+: a full-scale connected UNet for medical image segmentation[C]//2020 IEEE International Conference on Acoustics, Speech and Signal Processing. Barcelona: IEEE, 2020.
[18] ZHU Wenhui, CHEN Xiwen, QIU Peijie, et al. SelfReg-UNet: self-regularized UNet for medical image segmentation[C]//International Conference on Medical Image Computing and Computer Assisted Intervention. Morocco: MICCAI Society, 2024.
[19] MILLETARI F, NAVAB N, AHMADI S A. V-Net: fully convolutional neural networks for volumetric medical image segmentation[C]//2016 Fourth International Conference on 3D Vision. Stanford: IEEE, 2016.
[20] GIBSON E, GIGANTI F, HU Yipeng, et al. Automatic multi-organ segmentation on abdominal CT with dense V-networks[J]. IEEE transactions on medical imaging, 2018, 37(8): 1822-1834
[21] ROY S, KOEHLER G, ULRICH C, et al. MedNeXt: transformer-driven scaling of ConvNets for medical image segmentation[C]//International Conference on Medical Image Computing and Computer Assisted Intervention. Vancouver: MICCAI Society, 2023: 405-415.
[22] ZAGORUYKO S, KOMODAKIS N. Learning to compare image patches via cnvolutional neural networks[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015.
[23] HATAMIZADEH A, TANG Yucheng, NATH V, et al. UNETR: transformers for 3D medical image segmentation[C]//2022 IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa: IEEE, 2022.
[24] CAO Hu, WANG Yueyue, CHEN J, et al. Swin-UNet: UNet-like pure transformer for medical image segmentation[C]//European Conference on Computer Vision. Tel Aviv: ECVA, 2023.
[25] XU Qing, MA Zhicheng, NA He, et al. DCSAU-Net: a deeper and more compact split-attention U-Net for medical image segmentation[J]. Computers in biology and medicine, 2023, 154: 106626
[26] LIU Xueyu, SHI Guangze, WANG Rui, et al. Segment any tissue: one-shot reference guided training-free automatic point prompting for medical image segmentation[J]. Medical image analysis, 2025, 102: 103550
[27] RUAN Jiacheng, LI Jincheng, XIANG Suncheng. VM-UNet: vision mamba UNet for medical image segmentation[EB/OL]. (2024-02-04)[2025-04-20]. https://arxiv.org/abs/2402.02491.
[28] XING Zhaohu, YE Tian, YANG Yijun, et al. SegMamba: long-range sequential modeling mamba for 3D medical image segmentation[C]//International Conferenc on Medical Image Computing and Computer Assisted Intervention. Marrakesh: MICCAI Society, 2024.
[29] LIU Jiarun, YANG Hao, ZHOU Hongyu, et al. Swin-UMamba: mamba-based UNet with ImageNet-based pretraining[C]//International Conferenc on Medical Image Computing and Computer Assisted Intervention. MICCAI Society, 2024.
[30] LIU Mushui, DAN Jun, LU Ziqian, et al. CM-UNet: Hybrid CNN-Mamba UNet for remote sensing image semantic segmentation[EB/OL]. (2024-02-04)[2025-02-20]. https://arxiv.org/abs/2405.10530.
[31] CHI Lu, JIANG Borui, MU Yadong. Fast fourier convolution[J]. Advances in neural information processing systems, 2020, 33: 4479-4488
[32] ZHU Lei, WANG Xinjiang, KE Zhanghan, et al. BiFormer: vision transformer with bilevel routing attention[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver: IEEE, 2023.
[33] LEI Tao, SUN Rui, WANG Xuan, et al. CiT-net: convolutional neural networks hand in hand with vision transformers for medical image segmentation[C]//Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence. Maca: International Joint Conferences on Artificial Intelligence Organization, 2023.
[34] LI Chenxin, LIU Xinyu, LI Wuyang, et al. U-KAN makes strong backbone for medical image segmentation and generation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2025.
[35] LIU Yutong, ZHU Haijiang, LIU Mengting, et al. Rolling-UNet: revitalizing MLP’s ability to efficiently extract long-distance dependencies for medical image segmentation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver: AAAI, 2024.
[36] ZHANG Mingya, YU Yue, JIN Sun, et al. VM-UNET-V2: rethinking vision mamba UNet for medical image segmentation[C]//International Symposium on Bioinformatics Research and Applications. Kunming: CCF, 2024.
[37] WU Renkai, LIU Yinghao, NING Guochen, et al. UltraLight VM-UNet: parallel vision mamba significantly reduces parameters for skin lesion segmentation[J]. Patterns, 2025, 6(11): 101298
- 相似文献/References:
-
[1]程显毅,巩向普.改进的模糊C-均值算法在医学图像分割中的应用[J].智能系统学报,2010,5(1):80.
CHENG Xian-yi,GONG Xiang-pu.An improved fuzzy Cmeans algorithm for segmentation of medical images[J].CAAI Transactions on Intelligent Systems,2010,5():80.
[2]刘万军,姜岚,曲海成,等.融合CNN与Transformer的MRI脑肿瘤图像分割[J].智能系统学报,2024,19(4):1007.[doi:10.11992/tis.202301016]
LIU Wanjun,JIANG Lan,QU Haicheng,et al.MRI brain tumor image segmentation by fusing CNN and Transformer[J].CAAI Transactions on Intelligent Systems,2024,19():1007.[doi:10.11992/tis.202301016]
[3]孙锐,麦华煜,李徵,等.松弛分布一致性的半监督医学图像分割[J].智能系统学报,2026,21(1):132.[doi:10.11992/tis.202507034]
SUN Rui,MAI Huayu,LI Zhi,et al.Relaxed distribution-wise consistency for semi-supervised medical image segmentation[J].CAAI Transactions on Intelligent Systems,2026,21():132.[doi:10.11992/tis.202507034]
备注/Memo
收稿日期:2025-8-28。
基金项目:国家自然科学基金项目( 62271296；62201334)；陕西省创新能力支撑计划项目(2025RS-CXTD-012)；陕西省重点研发计划项目(2024GX-YBXM-121)；西安市中青年科技创新领军人才项目(25ZQRC00019)；陕西省教育厅科学研究计划项目(23JP014; 23JP022).
作者简介:彭晨阳，硕士研究生，主要研究方向为计算机视觉、机器学习。E-mail：2046591497@qq.com。;何立风，教授，博士，主要研究方向为数字图像处理、机器学习。 主持国家自然科学基金2项，发表学术论文60余篇，申请国家发明专利10余项。E-mail：helifeng@ist.aichipu.ac.jp。;雷涛，教授，博士，陕西科技大学电子信息与人工智能学院副院长，主要研究方向为人工智能、计算机视觉。主持国家自然基金5项，发表学术论文100余篇，其中12篇论文入选ESI高被引论文。E-mail：leitao@sust.edu.cn。
通讯作者:雷涛. E-mail：leitao@sust.edu.cn
更新日期/Last Update:
2026-01-05