Retinal Segmentation Method Based on Multi-Scale Bitemporal Fusion Module and GlobalGrouped Coordinate Attention
Liu Xuepeng1, Xu He1,2*, Ji Yimu1,2, Li Peng1,2
1(School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China) 2(Jiangsu HPC and Intelligent Processing Engineer Research Center, Nanjing 210023, China)
Abstract：Retinal vessel segmentation plays a crucial role in diagnosing retinal vascular diseases. Recent studies based on UNet have demonstrated promising performance in this task. However, segmenting thin retinal vessels with low contrast often leads to the loss of spatial information at certain stages. To address this issue, this paper proposed the novel network architecture, BFM-GGCA-UNet. The model employed convolutional kernels of different sizes at various stages to process feature maps, effectively capturing multi-scale features and forming a comprehensive feature representation. Furthermore, a multi-resolution convolutional interaction mechanism was introduced to expand the receptive field in both horizontal and vertical directions while maintaining the full image resolution. Additionally, to enhance this directional expansion, a global grouped attention mechanism was proposed. This mechanism leverages shared convolutional layers and an attention module to generate attention maps for the height and width dimensions, which weight the input feature maps to accentuate crucial features for a more precise prediction map. The proposed model was evaluated on five public datasets: DRIVE, STARE, CHASE_DB1, HRF and ARIA. The results showed that our model achieved AUCs of 98.89%, 99.50%, 99.13%,98.87%, and98.72%, and accuracies of97.25%, 98.06%,97.49%, 97.96% and 96.81% on these datasets, respectively, outperforming most existing methods. In conclusion, the designed BFM-GGCA-UNet effectively improved segmentation accuracy and delivered superior performance on these five fundamental retinal vessel segmentation datasets.
刘学鹏, 徐鹤, 季一木, 李鹏. 基于多尺度双时融合提取和全局分组注意力的视网膜分割方法[J]. 中国生物医学工程学报, 2026, 45(1): 11-24.
Liu Xuepeng, Xu He, Ji Yimu, Li Peng. Retinal Segmentation Method Based on Multi-Scale Bitemporal Fusion Module and GlobalGrouped Coordinate Attention. Chinese Journal of Biomedical Engineering, 2026, 45(1): 11-24.
[1] 王鹏竣,温莹. 糖尿病视网膜病变对糖尿病肾病的预测和诊断价值研究进展[J]. 眼科新进展,2025,45(2):151-155. [2] 周聪,代丽丽. 糖尿病性视网膜病与心血管风险的研究进展[J]. 心血管康复医学杂志,2023,32(5):520-523. [3] 汪有崧,裴峻鹏,李增辉,等. 深度学习的视网膜血管分割研究综述[J]. 计算机科学与探索,2024,18(8):1960-1978. [4] 黄晓鸣,何富运,唐晓虎,等. U-Net及其变体在医学图像分割中的应用研究综述[J]. 中国生物医学工程学报,2022,41(5):567-576. [5] Alom M, Taha T, Yakopcic C, et al. The history began from alexnet: A comprehensive survey on deep learning approaches[J]. https://arxiv.org/abs/1803.01164,2018-09-12/2025-03-28. [6] Wang Z, Gao X, Wu R, et al. Fully automatic image segmentation based on FCN and graph cuts[J]. Multimedia Systems,2022,28(5):1753-1765. [7] Kayalibay B, Jensen G, Smagt P. CNN-based segmentation of medical imaging data[J]. https://arxiv.org/abs/1701.03056,2017-07-25/2025-03-28. [8] Huang H, Lin L, Tong R, et al. Unet 3+: A full-scale connected unet for medical image segmentation[C] //Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Piscataway: IEEE, 2020:1055-1059. [9] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C] //Proceedings of the Medical Image Computing and Computer-assisted Intervention. Switzerland: Springer, 2015:234-241. [10] Azad R, Asadi-Aghbolaghi M, Fathy M, et al. Bi-directional ConvLSTM U-Net with densley connected convolutions[C] //Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops. Piscataway: IEEE, 2019: 406-415. [11] Zhang Q, Yang Y. Sa-net: Shuffle attention for deep convolutional neural networks[C] //Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Piscataway: IEEE, 2021:2235-2239. [12] Mou Lei, Zhao Yitian, Chen Li, et al. CS-Net: Channel and spatial attention network for curvilinear structure segmentation[C] //Proceedings of the Medical Image Computing and Computer Assisted Intervention, Springer. Switzerland: Springer, 2019:721-730. [13] Li Chunyang, Li Zhigang, Liu Weikang. TDCAU-Net: retinal vessel segmentation using transformer dilated convolutional attention-based U-Net method[J]. Physics in Medicine & Biology,2023,69(1):15003. [14] 季莉. 基于全分辨率注意力U-Net神经网络的区域分割方法[J]. 无线电工程,2023,53(9):1981-1989. [15] Staal J, Abr A, Niemeijer M, et al. Ridge-based vessel segmentation in color images of the retina[J]. IEEE Transactions on Medical Imaging,2004,23(4):501-509. [16] Carballal A, Novoa J, Fernandez-Lozano C, et al. Automatic multiscale vascular image segmentation algorithm for coronary angiography[J]. Biomedical Signal Processing and Control,2018,46:1-9. [17] Hoover A, Kouznetsova V, Goldbaum M. Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response[J]. IEEE Transactions on Medical Imaging,2000,19(3):203-210. [18] Budai A, Bock D, Maier A, et al. Robust vessel segmentation in fundus images[J]. International Journal of Biomedical Imaging, 2013,2013(1):154860. [19] Sim DA, Keane PA, Tufail A, et al. Automated retinal image analysis for diabetic retinopathy in telemedicine[J]. Current Diabetes Reports,2015,15:1-9. [20] Zhou Z, Rahman M, Tajbakhsh N, et al. Unet++: A nested u-net architecture for medical image segmentation[C] //Proceedings of the Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support.Granada: Springer International Publishing,2018:3-11. [21] 许洋,翟楠楠,倪维臻,等. 基于深度卷积和多层尺度特征融合的冠脉造影图像血管分割[J]. 中国生物医学工程学报,2025,44(1):34-42. [22] He Kaiming, Zhang Xiangyu, Ren Shaoqing, et al. Deep residual learning for image recognition[C] //Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016:770-778. [23] Oktay O, Schlemper J, Folgoc L, et al. Attention u-net: Learning where to look for the pancreas[J]. https://arxiv.org/abs/1804.03999,2018-05-20/2025-03-28. [24] 张子明,周庆华,薛洪省,等. 基于Unet+Attention的胸部CT影像支气管分割算法[J]. 中国生物医学工程学报,2024,43(1):60-69. [25] Guo Changlu, Szemenyei M, Yi Yugen, et al. Sa-unet: Spatial attention u-net for retinal vessel segmentation[C] //Proceedings of the 25th International Conference on Pattern Recognition (ICPR). Piscataway: IEEE,2021:1236-1242. [26] Seo S, Yoon H, Kim S, et al. Full-scale representation guided network for retinal vessel segmentation[J]. https://arxiv.org/abs/2501.18921,2025-01-31/2025-03-28.
