CADCNet:一种改进的视网膜血管分割算法  

CADCNet:An Improved Algorithm for Retinal Vessel Segmentation

作　　者：岳昱超 王迎美 秦嘉川 YUE Yu-chao;WANG Ying-mei;QIN Jia-chuan(School of Mathematics and Statistics,Shandong University of Technology,Zibo 255000,China;Shinva Medical Instrument Co.,Ltd.,Zibo 255086,China)

机构地区：[1]山东理工大学数学与统计学院,淄博255000 [2]山东新华医疗器械股份有限公司,淄博255086

出　　处：《科学技术与工程》2025年第3期962-968,共7页Science Technology and Engineering

基　　金：山东省自然科学基金面上项目(ZR2022MA027);中山大学广东省计算科学重点实验室开放基金(2021003);山东新华医疗器械股份有限公司横向课题(2D-C-20190158);山东省属普通本科高校教师访学研修项目。

摘　　要：目前传统的视网膜血管分割方法存在的视盘混淆引起的误分割、分割结果缺乏连续性,以及细节区域分割不精准等问题。为解决这一难题,提出了一种基于UNet的视网膜血管分割算法。该算法利用两个水平和垂直一维卷积和二维方形卷积的融合替代传统方形卷积,提高了眼球区域的表征能力;采用了多尺度分支增加特征空间的多样性,提升了网络的特征学习和表达能力。此外,为进一步改善分割效果,还将多层膨胀卷积引入自编码器的深层结构,替代了传统的简单池化操作,增大卷积核的大小,扩大了感受野范围,实现了多尺度浅层特征和深层特征信息的融合。本文算法在公开DRIVE和CHASE_DB1两个数据集上进行了评估,实验结果表明,本文算法的精确率和F_(1)上分别达到了0.9568、0.9598和0.8326、0.8304。与传统的UNet和近期部分UNet改进网络视网膜血管分割方法相比,本文算法在准确率、敏感度、特异性、F_(1)指标上表现出一定的优势,这一验证结果充分证明了本文所提出的模型在分割任务上具备较强的精确分割能力。Traditional retinal vessel segmentation methods often face challenges such as missegmentation caused by optic disc confusion,lack of continuity in segmentation results,and imprecise segmentation in detailed regions.To address these issues,a reti-nal vessel segmentation algorithm was proposed based on UNet.The algorithm replaced traditional square convolutions with a fusion of horizontal and vertical one-dimensional convolutions and two-dimensional square convolutions,enhancing the representation capa-bility of the eye region.A multi-scale branch approach was adopted to increase feature space diversity,thereby improving the net-work’s feature learning and expression capabilities.Additionally,to further enhance segmentation performance,multi-layer dilated convolutions was introduced into the deep structure of the autoencoder,replacing traditional simple pooling operations.This approach enlarged the convolution kernel size and expanded the receptive field,achieving a fusion of multi-scale shallow and deep feature in-formation.The proposed algorithm was evaluated on the public DRIVE and CHASE_DB1 datasets.Experimental results demonstrates that the algorithm achieves precision(0.9568 and 0.9598)and F_(1) scores(0.8326 and 0.8304),respectively.Compared with tra-ditional UNet and recent UNet-based retinal vessel segmentation methods,the proposed algorithm shows advantages in accuracy,sensitivity,specificity,and F_(1) metrics,these validation results fully demonstrate the proposed models strong capability in precise seg-mentation tasks.

关 键 词：视网膜血管分割 连续膨胀卷积 深度学习 不对称卷积 UNet模型 

分 类 号：R774[医药卫生—眼科]