A Deep Convolutional Encoder-Decoder Architecture for Retinal Blood Vessels Segmentation

Abstract

Over the last decades, various methods have been employed in medical images analysis. Some state-of-the-arts techniques such as deep learning have been recently applied to medical images analysis. This research proposes the application of deep learning technique in performing segmentation of retinal blood vessels. Analyzing and segmentation of retina vessels has assisted in diagnosis and monitoring of some diseases. Diseases such as age-related fovea degeneration, diabetic retinopathy, glaucoma, hypertension, arteriosclerosis and choroidal neovascularization can be effectively managed by the analysis of retinal vessels images. In this work, a Deep Convolutional Encoder-Decoder Architecture for the segmentation of retinal vessels images is proposed. The proposed method is a deep learning system composed of an encoder and decoder mechanism allows a low resolution image set of retinal vessels to be analyzed by set of convolutional layers in the encoder unit before been sent into a decoder unit for final segmented output. The proposed system was evaluated using some evaluation metrics such as dice coefficient, jaccard index and mean of intersection. The review of the existing works was also carried out. It could be shown that the proposed system outperforms many existing methods in the segmentation of retinal vessels images.