Retinal Blood Vessel Segmentation using odd heavy U-net Architecture

Electronics and Communication Engineering

Blood vessel segmentation plays a vital role in computer aided diagnosis and treatment of retinal diseases. This is the reason why blood vessel segmentation has gained wide popularity among researchers. In this project we implement blood vessel segmentation based on an improved Odd heavy U-NET convolutional neural network (CNN) architecture. The architecture is very similar to U-net architecture only even layers have three convolutions followed by ReLU whereas odd layers have two convolutions followed by ReLU. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. Multiscale input layer and dense blocks are introduced into the conventional U-NET, so that the network can make use richer spatial context information. Especially for thin blood vessels, which are difficult to detect because of their low contrast with the background pixels, this segmentation results have been improved. This is the simplest architecture used for recognition of various retinal diseases. We show that such a network can be trained end to end from very few images and performs the prior best method. Further, the segmented outputs were able to cover thinner blood vessels better than previous methods aiding in early detection of pathologies.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.