摘 要：彩色眼底图像的视网膜血管分析可以帮助医生诊断许多眼科和全身性疾病，具有十分重要的临床意义。为进一步提高视网膜血管的分割效果，文章提出一个基于注意力 U-Net 网络的视网膜血管分割方法，该方法使用 U-Net 结合通道注意力机制以提高分割准确率，在公开数据集 DRIVE 的灵敏度、特异性和准确率分别为 0.772 6，0.984 7 和 0.966 0，优于现有的许多方法。
关键词：注意力；U-Net；血管分割
DOI:10.19850/j.cnki.2096-4706.2023.06.017
基金项目：淮安市创新服务能力建设计划 -重点实验室建设项目（HAP201904）
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2023）06-0065-04
Segmentation of Retinal Vascular in Color Fundus Images Based on Attention U-Net Network
ZHENG Liping, ZHANG Tianshu, CAO Keyin
(Jiangsu Vocational College of Electronics and Information, Huaian 223001, China)
Abstract: Retinal vascular analysis of color fundus images can help doctors diagnose many ophthalmic and systemic diseases, which is of great clinical significance. To further improve the segmentation effect of retinal vascular, this paper proposes a retinal vascular segmentation method based on attention U-Net network, which uses U-Net and channel attention mechanism to improve segmentation accuracy, The sensitivity, specificity and accuracy of DRIVE in the open dataset are 0.772 6, 0.984 7 and 0.966 0 respectively, which are superior to many existing methods.
Keywords: attention; U-Net; vascular segmentation
参考文献：
[1] 朱承璋，邹北骥，向遥，等 . 彩色眼底图像视网膜血管分割方法研究进展 [J]. 计算机辅助设计与图形学学报，2015，27（11）：2046-2057.
[2] BAKER M L，HAND P J，WANG J J，et al. Retinal Signs and Stroke：Revisiting the Link Between the Eye and Brain [J].Stroke， 2008，39：1371-1379.
[3] ZHANG Y P，HSU W，LEE M L. Detection of Retinal Blood Vessels Based on Nonlinear Projections [J].Journal of Signal Processing Systems，2009，55：103-112.
[4] FRAZ M M，BARMAN S A，REMAGNINO P，et al. An approach to localize the retinal blood vessels using bit planes and centerline detection [J].Computer Methods Programs in Biomedicine， 2012，108（2）：600-616.
[5] ALOM M Z，YAKOPCIC C，HASAN M，et al. Recurrent residual U-Net for medical image segmentation [J/OL].Journal of Medical Imaging，2019，6（1）[2022-10-18].https://www. zhangqiaokeyan.com/academic-journal-foreign-pmc_journal-medicalimaging_thesis/040006256846.html.
[6] RONNEBERGER O，FISCHER P，BROX T. U-Net： Convolutional Networks for Biomedical Image Segmentation [J/ OL].arXiv:1505.04597 [cs.CV].[2022-10-19].https://arxiv.org/ abs/1505.04597.
[7] SOOMRO T A，AFIFI A J，GAO J B，et al. Boosting Sensitivity of a Retinal Vessel Segmentation Algorithm with Convolutional Neural Network [C]//2017 International Conference on Digital Image Computing：Techniques and Applications （DICTA）.Sydney：IEEE，2017：1-8.
[8] 孟娴静，李菁，吕晓琪，等 . 基于改进 SegNet 的眼底图像血管分割 [J]. 计算机工程与设计，2020，41（11）：3148-3153.
[9] BADRINARAYANAN V，KENDALL A，CIPOLLA R. SegNet：A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2017，39（12）：2481-2495.
[10] 陈林，陈亚军，沈锐 . 基于残差卷积注意力网络的视网膜血管分割算法 [J]. 内江师范学院学报，2021，36（2）：49-55.
[11] 邓小波，刘奇，陈曦，等 . 基于通道注意力双路径架构网络分割视网膜血管 [J]. 中国医学影像技术，2021，37（10）：1543-1547.
[12] STAAL J，ABRÀMOFF M D，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.
[13] NIEMEIJER M，STAAL J，GINNEKEN B V，et al. Comparative study of retinal vessel segmentation methods on a new publicly available database [EB/OL].[2022-10-18].https:// www.spiedigitallibrary.org/conference-proceedings-of-spie/5370/1/ Comparative-study-of-retinal-vessel-segmentation-methods-on-anew/10.1117/12.535349.short.
[14] ALOM M Z，HASAN M，YAKOPCIC C，et al. Recurrent Residual Convolutional Neural Network based on U-Net （R2U-Net） for Medical Image Segmentation [J/OL].arXiv:1802.06955 [cs.CV]. [2022-10-18].https://arxiv.org/abs/1802.06955v2.
[15] LISKOWSKI P，KRAWIEC K. Segmenting Retinal Blood Vessels With Deep Neural Networks [J].IEEE Transactions on MedicalImaging，2016，35（11）：2369-2380.
[16] LI Q L，FENG B W，XIE L P，et al. A Cross-Modality Learning Approach for Vessel Segmentation in Retinal Images [J].IEEE Transactions on Medical Imaging，2016，35（1）：109-118.
[17] YAN Z Q，YANG X，CHENG K T. Joint Segment-Level and Pixel-Wise Losses for Deep Learning Based Retinal Vessel Segmentation [J].IEEE Transactions on Biomedical Engineering，2018，65（9）： 1912-1923.
作者简介：郑丽萍（1980—），女，汉族，河南开封人，讲师，工程师，硕士，研究方向：计算机视觉等；张天舒（1995—），男，汉族，河南商丘人，助教，硕士，研究方向：计算机视觉；曹珂崯（1994—），男，汉族，山东菏泽人，助教，硕士，研究方向：数据分析。