Dense Residual Convolutional Auto Encoder For Retinal Blood Vessels Segmentation

Abstract

In order to overcome the difficulties in retinal blood vessel segmentation and aid ophthalmologists in diagnosis of diabetic retinopathy and glaucoma, there is a need for effective segmentation techniques. One such efficient technique is to use a model for segmentation using deep learning. In this paper, an auto encoder deep learning network model based on residual path and U-net has been implemented to effectively segment the retinal blood vessels. Our network model has been implemented and tested on DRIVE dataset. This proposed model is reporting an increase in efficiency and Area under ROC compared to previous methods.