Preliminary Study on Deep-learning for Retinal Vessels Segmentation

Abstract

The retina is an internal part of the human eye that plays a crucial role in vision. Retinal vessel segmentation is an important basis for the identification and classification of Diabetic Retinopathy (DR), Age related macular degeneration (AMD)，Retinal Detachment, and other retinal diseases. At the same time, the morphological structure of retinal blood vessels can also be used to diagnose cardiovascular diseases, which is important for early diagnosis and prevention of exacerbation of the disease. There are already advanced state-of-the-art methods to help automatically segment and identify retinal vessel diseases. This paper presents the methods of the principle and applications of deep learning in retinal image analysis. We described various segmentation methods based on deep learning. Analyzing along with the limitations of each method. Finally, we proposed some suggestions for the improvement of retinal image analysis.