Abstract—The state of blood vessels in the retina is an important factor used to diagnose the presence of several eye diseases and heart conditions. This is the reason why blood vessels segmentation of fundus images has gained wide popularity among researchers. This paper proposes blood vessel segmentation method based on an improved U-net Convolutional Neural Network (CNN) architecture. The proposed method involves very minimal fundus pre-processing and no post-processing of segmented blood vessel, thus, making the method very simple and easy to use. The method is tested on four publicly available databases which are: the DRIVE, CHASE _DB1, STARE and the HRF databases. The proposed method outperforms existing methods in accuracy, specificity and sensitivity on the four databases. A sensitivity of 0.8309, 0.7796, 0.7506, 0.8059 and a specificity of 0.9742, 0.9864, 0.9824 and 0.9826 are achieved on the DRIVE, CHASE_DB1, STARE and HRF databases respectively.

Index Terms—segmentation, blood vessel, convolutional neural network, u-net, retina fundus image

Cite: Afolabi O. Joshua, Fulufhelo V. Nelwamondo, and Gugulethu Mabuza-Hocquet, "Blood Vessel Segmentation from Fundus Images Using Modified U-net Convolutional Neural Network," Journal of Image and Graphics, Vol. 8, No. 1, pp. 21-25, March 2020. doi: 10.18178/joig.8.1.21-25

Copyright © 2020 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.