基于PCNN和生成对抗网络的视网膜血管分割方法  

Segmentation of Retinal Vessels Based on PCNN and GAN

作　　者：罗耀 LUO Yao(Guangdong Mechanical&Electrical Polytechnic School of Computer Science and Design,Guangzhou 510550 China)

机构地区：[1]广东机电职业技术学院人工智能学院,广东广州510550

出　　处：《自动化技术与应用》2025年第1期158-162,共5页Techniques of Automation and Applications

基　　金：广东省特色创新项目(2022KTSCX236)。

摘　　要：视网膜血管分割能为高血压、糖尿病以及青光眼等眼科疾病的确诊和治疗提供支持。利用计算机辅助技术的视网膜血管分割相比人工诊断更容易减少漏诊和误诊的机会。为提高视网膜血管分割准确率,提出基于PCNN和生成对抗网络的视网膜分割的方法。将PCNN和生成对抗网络相结合,将生成对抗网络中的判别器嵌入PCNN为主体的循环结构中。生成对抗网络的博弈竞争机制赋予判别器不断自我升级的动力和活力,而不断改进的判别器为PCNN提供了分割的标准。最后对提出的方法进行定性和定量的分析,实验结果表明其在视网膜血管分割方面具有比较好的表现。Retinal vascular segmentation can provide support for the diagnosis and treatment of ocular diseases such as hypertension,diabetes and glaucoma.Compared with manual diagnosis,computer-aided retinal vascular segmentation is easier to reduce the chance of missed diagnosis and misdiagnosis.In order to improve the accuracy of retinal vascular segmentation,a method based on PCNN and GAN is proposed.It combines PCNN with GAN,and embedding the discriminant in the GAN into the loop structure with PCNN as the main body.The game competition mechanism against the network gives the discriminant the power and vitality of continuous self-upgrading,and the constantly improved discriminant provides the segmentation standard for PCNN.Finally,the experimental results show that the proposed method has a good performance in the segmentation of retinal vessels.

关 键 词：血管分割 图像处理 PCNN 生成对抗网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]