双路自优化小样本医学图像分割网络

doi:10.3778/j.issn.1002-8331.2409-0246

摘要/Abstract

摘要： 当前，小样本医学图像分割方法主要通过特征空间对齐和增强模型泛化性以指导查询集分割，但其忽略了支持集与查询集病理器官间存在显著类内差异以及前景与背景之间存在边缘误差。提出一种双路自优化小样本医学图像分割网络（dual-branch self-optimizing segmentation network，DSSNet），该网络充分利用医学图像的结构位置先验，采用前景与背景双路径并行优化原型策略，以提升对样本边缘分割精度。同时提出了一种自匹配的动态优化策略，通过利用高置信度原型进行原型校准，缓解支持样本与查询样本间的类内差异。该策略通过在双分支网络中引入前景与背景的并行优化，增强了两者之间的结构信息互补。此外，提出了一种多尺度特征聚合模块，通过聚合模型中不同尺度的语义特征，在保留更多细节信息的同时增强模型的长距离相关性依赖。模型在公开的医学图像分割数据集Abd-MRI和Card-MRI上进行了充分的对比实验与消融实验，结果验证了所提模型与当前最先进的小样本医学图像分割方法相比，在分割准确率上展现出了显著的性能优势。

关键词: 小样本学习, 医学图像分割, 注意力机制, 原型自校准

Abstract: Currently, few-shot medical image segmentation methods primarily guide query set segmentation through feature space alignment and enhanced model generalization. However, they overlook significant intra-class differences between pathological organs in support and query sets, as well as edge errors between foreground and background. This paper introduces a dual-branch self-optimizing segmentation network (DSSNet) that leverages structural and positional priors of medical images. DSSNet employs parallel optimization of foreground and background pathways to improve edge segmentation accuracy. Additionally, a self-matching dynamic optimization strategy is proposed to calibrate prototypes using high-confidence samples, mitigating intra-class discrepancies between support and query sets. This strategy enhances complementary structural information through parallel foreground and background optimization within the dual-branch network. A multi-scale feature aggregation module is also presented, which aggregates semantic features at various scales to strengthen long-range dependencies while retaining detailed information. Extensive comparative experiments and ablation studies on the Abd-MRI and Card-MRI datasets demonstrate that the proposed model significantly outperforms current state-of-the-art few-shot medical image segmentation methods in segmentation accuracy.

Key words: few-shot learining, medical image segmentation, attention mechanism, prototype self-refinement

高峰, 降爱莲, 冀伟, 肖慈美. 双路自优化小样本医学图像分割网络[J]. 计算机工程与应用, 2025, 61(24): 228-239.

GAO Feng, JIANG Ailian, JI Wei, XIAO Cimei. Dual-Branch Self-Optimizing Segmentation Network for Few-Shot Medical Image Segmentation[J]. Computer Engineering and Applications, 2025, 61(24): 228-239.

参考文献

[1] SHABAN A, BANSAL S, LIU Z, et al. One-shot learning for semantic segmentation[J]. arXiv:1709.03410, 2017.
[2] RAKELLY K , SHELHAMER E, DARRELL T, et al. Conditional networks for few-shot semantic segmentation[C]//Proceedings of the International Conference on Learning Representations, 2018.
[3] OUYANG C, BIFFI C, CHEN C, et al. Self-supervision with superpixels: training few-shot medical image segmentation without annotation[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2020: 762-780.
[4] HUANG Y, LIU J M, CHEN H. Self-reinforcing for few-shot medical image segmentation[C]//Proceedings of the 2023 IEEE International Conference on Image Processing. Piscataway: IEEE, 2023: 655-659.
[5] SHEN Q Q, LI Y N, JIN J Y, et al. Q-net: query-informed few-shot medical image segmentation[C]//Proceedings of SAI Intelligent Systems and Applications. Cham: Springer, 2024: 610-628.
[6] ZHU Y Z, WANG S D, XIN T, et al. Few-shot medical image segmentation via a region-enhanced prototypical transformer[C]//Proceedings of the International Conference on Medical Image Computing and Computer Assisted Intervention. Cham: Springer, 2023: 271-280.
[7] ZHU Y Z, WANG S D, XIN T, et al. Partition-A-medical-image: extracting multiple representative subregions for few-shot medical image segmentation[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 5016312.
[8] CHRIST P F, ETTLINGER F, GRüN F, et al. Automatic liver and tumor segmentation of CT and MRI volumes using cascaded fully convolutional neural networks[J]. arXiv:1702. 05970, 2017.
[9] HUANG W D, XIAO B, HU J W, et al. Location-aware transformer network for few-shot medical image segmentation[C]//Proceedings of the 2023 IEEE International Conference on Bioinformatics and Biomedicine. Piscataway: IEEE, 2024: 1150-1157.
[10] TIAN Z T, ZHAO H S, SHU M, et al. Prior guided feature enrichment network for few-shot segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(2): 1050-1065.
[11] ZHANG B F, XIAO J M, QIN T. Self-guided and cross-guided learning for few-shot segmentation[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 8308-8317.
[12] LI G, JAMPANI V, SEVILLA-LARA L, et al. Adaptive prototype learning and allocation for few-shot segmentation[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 8330-8339.
[13] LANG C B, CHENG G, TU B F, et al. Learning what not to segment: a new perspective on few-shot segmentation[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 8047-8057.
[14] ZHANG X L, WEI Y C, YANG Y, et al. SG-one: similarity guidance network for one-shot semantic segmentation[J]. IEEE Transactions on Cybernetics, 2020, 50(9): 3855-3865.
[15] WANG K X, LIEW J H, ZOU Y T, et al. PANet: few-shot image semantic segmentation with prototype alignment[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 9196-9205.
[16] CAO L L, GUO Y B, YUAN Y, et al. Prototype as query for few shot semantic segmentation[J]. arXiv:2211.14764, 2022.
[17] XU Q X, ZHAO W T, LIN G S, et al. Self-calibrated cross attention network for few-shot segmentation[C]//Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2024: 655-665.
[18] PENG B H, TIAN Z T, WU X Y, et al. Hierarchical dense correlation distillation for few-shot segmentation[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 23641-23651.
[19] MIN J H, KANG D, CHO M. Hypercorrelation squeeze for few-shot segmenation[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2022: 6921-6932.
[20] LIU H F, PENG P, CHEN T, et al. FECANet: boosting few-shot semantic segmentation with feature-enhanced context-aware network[J]. IEEE Transactions on Multimedia, 2023, 25: 8580-8592.
[21] GUHA ROY A, SIDDIQUI S, P?LSTERL S, et al. ‘Squeeze & excite’ guided few-shot segmentation of volumetric images[J]. Medical Image Analysis, 2020, 59: 101587.
[22] FENG R W, ZHENG X S, GAO T X, et al. Interactive few-shot learning: limited supervision, better medical image segmentation[J]. IEEE Transactions on Medical Imaging, 2021, 40(10): 2575-2588.
[23] SUN L Y, LI C X, DING X H, et al. Few-shot medical image segmentation using a global correlation network with discriminative embedding[J]. Computers in Biology and Medicine, 2022, 140: 105067.
[24] HANSEN S, GAUTAM S, JENSSEN R, et al. Anomaly detection-inspired few-shot medical image segmentation through self-supervision with supervoxels[J]. Medical Image Analysis, 2022, 78: 102385.
[25] LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft COCO: common objects in context[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2014: 740-755.
[26] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 770-778.
[27] ZHANG Q L, YANG Y B. ResT: an efficient transformer for visual recognition[J]. arXiv:2105.13677, 2021.
[28] AURENHAMMER F. Voronoi diagrams: a survey of a fundamental geometric data structure[J]. ACM Computing Surveys, 1991, 23(3): 345-405.
[29] CHENG Z M, WANG S D, XIN T, et al. Few-shot medical image segmentation via generating multiple representative descriptors[J]. IEEE Transactions on Medical Imaging, 2024, 43(6): 2202-2214.
[30] KAVUR A E, GEZER N S, BAR?? M, et al. CHAOS challenge-combined (CT-MR) healthy abdominal organ segmentation[J]. Medical Image Analysis, 2021, 69: 101950.
[31] ZHUANG X H. Multivariate mixture model for myocardial segmentation combining multi-source images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(12): 2933-2946.
[32] WU H S, XIAO F Y, LIANG C X. Dual contrastive learning withAnatomical auxiliary supervision for few-shot medical image segmentation[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2022: 417-434.
[33] SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization[J]. International Journal of Computer Vision, 2020, 128(2): 336-359.