基于可形变编码器-解码器网络的视网膜血管精准分割

Deformable encoder-decoder network for accurate retinal vessel segmentation

摘要

针对现有方法在视网膜血管分割中，捕获血管不规则形态及多尺度特征融合方面的不足，提出一种轻量化可形变的编码器-解码器视网膜血管分割网络ResDC-Net，该网络由编码器、跳跃连接模块和解码器组成。编码器和解码器部分主要由残差可形变卷积模块构成，该模块利用可形变卷积动态适配血管的形态，并通过残差连接增强特征流，有效解决传统卷积对弯曲血管、分支结构的表征不足问题，提升血管形变的建模能力。运用分组聚合桥接模块作为跳跃连接模块，该模块融合编码器输出的不同深度的信息和解码器在每个阶段生成的掩码信息，强化语义信息与空间细节的互补性，有助于恢复血管边缘细节。在STARE、CHASE_DB1、HRF（分辨率为876×584、1 752×1 168）公开数据集上进行实验，其中特异性分别达到98.66%、98.67%、98.43%和98.42%，准确率分别达到97.47%、97.63%、97.18%和97.17%。实验表明本文网络在视网膜血管分割任务中的有效性、轻量性与优越性。

Abstract

To overcome the drawbacks of existing approaches in capturing irregular vascular morphologies and integrating multi-scale features, this study proposes a lightweight deformable encoder-decoder network named ResDC-Net for retinal vessel segmentation. The network consists of 3 components: an encoder, skip connection modules, and a decoder. Both the encoder and decoder are mainly composed of residual deformable convolution modules. These modules utilize deformable convolutions to dynamically adapt to vascular morphology and enhance feature propagation through residual connections, which effectively compensates for the limited representation capacity of traditional convolutions for curved vessels and branches, thereby improving the modeling for vascular deformations. A group aggregation bridge module serves as the skip connection module. This module integrates multi-depth information output by the encoder and the mask information generated by the decoder at each stage, strengthening the complementarity between semantic information and spatial details, which is conducive to restoring vessel edge details. Experiments on the public STARE, CHASE_DB1, and HRF (resolution: 876×584, 1752×1168) datasets show that the proposed method achieves specificity of 98.66%, 98.67%, 98.43% and 98.42%, and accuracy of 97.47%, 97.63%, 97.18% and 97.17%, respectively, demonstrating its effectiveness, lightweight property, and superiority for retinal vessel segmentation.

关键词

医学图像 / 视网膜血管分割 / 残差可形变卷积 / 分组聚合桥接 / 轻量化网络

Key words

medical image / retinal vessel segmentation / residual deformable convolution / group aggregation bridge / lightweight network

参考文献

[1] Hassan G, El—Bendary N, Hassanien AE, et al. Retinal blood vessel segmentation approach based on mathematical morphology[J]. Procedia Comput Sci, 2015, 65: 612-622.

[2] Smart TJ, Richards CJ, Bhatnagar R, et al. A study of red blood cell deformability in diabetic retinopathy using optical tweezers[C]// Optical Trapping and Optical Micromanipulation XII. Bellingham, WA, USA: SPIE, 2015: 954825.

[3] Laibacher T, Weyde T, Jalali S . M2U—Net: effective and efficient retinal vessel segmentation for resource—constrained environments[J]. (2019—04—23). https://arxiv.org/abs/1811.07738.

[4] Cheung CY, Zheng YF, Hsu W, et al. Retinal vascular tortuosity, blood pressure, and cardiovascular risk factors[J]. Ophthalmology, 2011, 118(5): 812-818.

[5] Yang YX, Sun LX, Tang ZW, et al. A precise image—based retinal blood vessel segmentation method using TAOD—CFNet[J]. Biomed Signal Process Control, 2025, 107: 107815.

[6] 李天培, 陈黎 . 基于双注意力编码—解码器架构的视网膜血管分割[J]. 计算机科学, 2020, 47(5): 166-171.

[7] Li TP, Chen L . Retinal vessel segmentation based on dual attention and encoder—decoder structure[J]. Computer Science, 2020, 47(5): 166-171.

[8] 梁礼明, 余洁, 周珑颂, 等 . 多尺度密集注意力网络用于视网膜血管分割[J]. 激光与光电子学进展, 2023, 60(6): 112-121.

[9] Liang LM, Yu J, Zhou LS, et al. Multiscale dense attention network for retinal vessel segmentation[J]. Laser & Optoelectronics Progress, 2023, 60(6): 112-121.

[10] Ronneberger O, Fischer P, Brox T . U—Net: convolutional networks for biomedical image segmentation[C]// Medical Image Computing and Computer—Assisted Intervention—MICCAI 2015. Cham: Springer International Publishing, 2015: 234-241.

[11] Li LZ, Verma M, Nakashima Y, et al. IterNet: retinal image segmentation utilizing structural redundancy in vessel networks[C]// 2020 IEEE Winter Conference on Applications of Computer Vision (WACV). Piscataway, NJ, USA: IEEE, 2020: 3645-3654.

[12] Yang B, Qin L, Peng H, et al. SDDC—Net: a U—shaped deep spiking neural P convolutional network for retinal vessel segmentation[J]. Digit Signal Process, 2023, 136: 104002.

[13] Wang ZY, Jia LV, Liang HC. Partial class activation mapping guided graph convolution cascaded U—Net for retinal vessel segmentation[J]. Comput Biol Med, 2024, 178: 108736.

[14] 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.

[15] Wang YS, Pei JP, Li ZH, et al. Review of research on deep learning in retinal blood vessel segmentation[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(8): 1960-1978.

[16] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, NY, USA. Curran Associates Inc., 2017: 6000-6010.

[17] Xiong YW, Li ZQ, Chen YT, et al. Efficient deformable ConvNets: rethinking dynamic and sparse operator for vision applications[C]// 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2024: 5652-5661.

[18] Kang N, Wang MF, Pang C, et al. Cross—patch feature interactive net with edge refinement for retinal vessel segmentation[J]. Comput Biol Med, 2024, 174: 108443.

[19] Fraz MM, Remagnino P, Hoppe A, et al. An ensemble classification—based approach applied to retinal blood vessel segmentation[J]. IEEE Trans Biomed Eng, 2012, 59(9): 2538-2548.

[20] Budai A, Bock R, Maier A, et al. Robust vessel segmentation in fundus images[J]. Int J Biomed Imaging, 2013, 2013: 154860.

[21] Han Z, Yin YL, Meng XJ, et al. Blood vessel segmentation in pathological retinal image[C]// 2014 IEEE International Conference on Data Mining Workshop. Piscataway, NJ, USA: IEEE, 2014: 960-967.

[22] Wang CL, Oda M, Hayashi Y, et al. Tensor—cut: a tensor—based graph—cut blood vessel segmentation method and its application to renal artery segmentation[J]. Med Image Anal, 2020, 60: 101623.

[23] Kalaie S, Gooya A . Vascular tree tracking and bifurcation points detection in retinal images using a hierarchical probabilistic model[J]. Comput Methods Programs Biomed, 2017, 151: 139-149.

[24] Sun JD, Peng YJ, Guo YF, et al. Segmentation of the multimodal brain tumor image used the multi—pathway architecture method based on 3D FCN[J]. Neurocomputing, 2021, 423: 34-45.

[25] Zhou ZW, Rahman Siddiquee MM, Tajbakhsh N, et al. UNet++: a nested U—Net architecture for medical image segmentation[C]// Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. Cham: Springer International Publishing, 2018: 3-11.

[26] Huang HM, Lin LF, Tong RF, et al. UNet 3+: a full—scale connected UNet for medical image segmentation[C]// ICASSP 2020—2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Piscataway, NJ, USA: IEEE, 2020: 1055-1059.

[27] Gu ZW, Cheng J, Fu HZ, et al. CE—Net: context encoder network for 2D medical image segmentation[J]. IEEE Trans Med Imaging, 2019, 38(10): 2281-2292.

[28] Guo CL, Szemenyei M, Yi YG, et al. SA—UNet: spatial attention U—Net for retinal vessel segmentation[C]// 2020 25th International Conference on Pattern Recognition (ICPR). Piscataway, NJ, USA: IEEE, 2021: 1236-1242.

[29] Guo CL, Szemenyei M, Pei Y, et al. SD—Unet: a structured dropout U—Net for retinal vessel segmentation[C]// 2019 IEEE 19th International Conference on Bioinformatics and Bioengineering (BIBE). Piscataway, NJ, USA: IEEE, 2019: 439-444.

[30] Mou L, Zhao YT, Fu HZ, et al. CS2—Net: deep learning segmentation of curvilinear structures in medical imaging[J]. Med Image Anal, 2021, 67: 101874.

[31] Liu WT, Yang HH, Tian T, et al. Full—resolution network and dual—threshold iteration for retinal vessel and coronary angiograph segmentation[J]. IEEE J Biomed Health Inform, 2022, 26(9): 4623-4634.

[32] Gao G, Li JY, Yang L, et al. A multi—scale global attention network for blood vessel segmentation from fundus images[J]. Measurement, 2023, 222: 113553.

[33] Qu ZW, Zhuo L, Cao J, et al. TP—Net: two—path network for retinal vessel segmentation[J]. IEEE J Biomed Health Inform, 2023, 27(4): 1979-1990.

[34] Xiang Z, Ning C, Li M, et al. AFFD—Net: a dual—decoder network based on attention—enhancing and feature fusion for retinal vessel segmentation[J]. IEEE Access, 2023, 11: 45871-45887.

[35] Zhou W, Bai WQ, Ji JH, et al. Dual—path multi—scale context dense aggregation network for retinal vessel segmentation[J]. Comput Biol Med, 2023, 164: 107269.

[36] Huang H, Shang ZH, Yu CH . FRD—Net: a full—resolution dilated convolution network for retinal vessel segmentation[J]. Biomed Opt Express, 2024, 15(5): 3344-3365.

[37] Wu HS, Wang W, Zhong JF, et al. SCS—Net: a scale and context sensitive network for retinal vessel segmentation[J]. Med Image Anal, 2021, 70: 102025.

[38] Xie Y, Shang JK, Yang Q, et al. ARSA—UNet: atrous residual network based on Structure—Adaptive model for retinal vessel segmentation[J]. Biomed Signal Process Control, 2024, 96, Part B: 106595.

[39] Wang JH, Li XB, Ma ZD . Multi—scale three—path network (MSTP—Net): a new architecture for retinal vessel segmentation[J]. Measurement, 2025, 250: 117100.

[40] Ruan JC, Xiang SC, Xie MY, et al. MALUNet: a multi—attention and light—weight UNet for skin lesion segmentation[C]// 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). Piscataway, NJ, USA: IEEE, 2022: 1150-1156.

[41] Ruan JC, Xie MY, Gao JS, et al. EGE—UNet: an efficient group enhanced UNet for skin lesion segmentation[C]// Medical Image Computing and Computer Assisted Intervention—MICCAI 2023. Cham: Springer Nature Switzerland, 2023: 481-490.

[42] Akgül İ. A pooling method developed for use in convolutional neural networks[J]. Comput Model Eng Sci, 2024, 141(1): 751-770.

[43] Milletari F, Navab N, Ahmadi SA . V—Net: fully convolutional neural networks for volumetric medical image segmentation[C]// 2016 Fourth International Conference on 3D Vision (3DV). Piscataway, NJ, USA: IEEE, 2016: 565-571.

[44] Liu YH, Shen J, Yang L, et al. Wave—Net: a lightweight deep network for retinal vessel segmentation from fundus images[J]. Comput Biol Med, 2023, 152: 106341.

[45] Deng XY, Ye JH . A retinal blood vessel segmentation based on improved D—MNet and pulse—coupled neural network[J]. Biomed Signal Process Control, 2022, 73: 103467.

[46] Liu MT, Wang YY, Wang L, et al. IMFF—Net: an integrated multi—scale feature fusion network for accurate retinal vessel segmentation from fundus images[J]. Biomed Signal Process Control, 2024, 91: 105980.

[47] Panchal S, Kokare M . ResMU—Net: residual multi—kernel U—Net for blood vessel segmentation in retinal fundus images[J]. Biomed Signal Process Control, 2024, 90: 105859.

[48] Zhang HB, Zhong X, Li ZJ, et al. TiM—Net: transformer in M—Net for retinal vessel segmentation[J]. J Healthc Eng, 2022, 2022: 9016401.

[49] Li Y, Zhang Y, Liu JY, et al. Global transformer and dual local attention network via deep—shallow hierarchical feature fusion for retinal vessel segmentation [J]. IEEE Trans Cybern, 2023, 53(9): 5826-5839.

[50] Tong L, Li TJ, Zhang Q, et al. LiViT—Net: a U—Net—like, lightweight transformer network for retinal vessel segmentation[J]. Comput Struct Biotechnol J, 2024, 24: 213-224.

基金资助

国家自然科学基金(61871055)