Retinal blood vessel segmentation based on Densely Connected U-Net

Mathematical Biosciences and Engineering

卷:17

Mengnan Ma¹  Yinlin Cheng¹  Liangjun Zhang²  ChenJin Jin³  Li Ma³  Yi Zhou⁴ 

[1] 1. School of Biomedical Engineering, Sun Yat-sen University, Guangzhou 510006, China3. Department of Medical Informatics, Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou 510006, China;

[2] 1. School of Biomedical Engineering, Sun Yat-sen University, Guangzhou 510006, China;

[3] 2. Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangzhou 510006, China;

[4] 3. Department of Medical Informatics, Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou 510006, China;

关键词: retinal fundus image;    u-net;    blood vessel segmentation;    neural networks;    dense block;   

DOI  :  10.3934/mbe.2020175

来源: DOAJ

【 摘 要 】

The segmentation of blood vessels from retinal images is an important and challenging task in medical analysis and diagnosis. This paper proposes a new architecture of the U-Net network for retinal blood vessel segmentation. Adding dense block to U-Net network makes each layer’s input come from the all previous layer’s output which improves the segmentation accuracy of small blood vessels. The effectiveness of the proposed method has been evaluated on two public datasets (DRIVE and CHASE_DB1). The obtained results (DRIVE： Acc = 0.9559， AUC = 0.9793， CHASE_DB1： Acc = 0.9488， AUC = 0.9785) demonstrate the better performance of the proposed method compared to the state-of-the-art methods. Also, the results show that our method achieves better results for the segmentation of small blood vessels and can be helpful to evaluate related ophthalmic diseases.

【 授权许可】

Unknown