Retinal Vessel Segmentation using Deep Learning – A Study

The Retinal Vasculature conditions are important and reliable biomarkers for several cardiovascular and ophthalmologic diseases including Retinopathy of Prematurity, Diabetic Retinopathy, Glaucoma, Macular Degeneration, etc. The retinal-related diseases can be detected and diagnosed in the early stage before developing complexities and causing loss of sight by thoroughly understanding the vasculature condition. Analyzing these conditions is time consuming and tedious task. There have been many studies conducted to automatically extract retinal vessel information from fundus image and this is called retinal vessel segmentation. This paper provides a survey of the existing work that had been conducted to achieve this objective using Deep Learning methods. Deep Learning has become the state-of-the-art for solving computer vision problems. This paper briefly discusses U-Net based network architectures proposed for object segmentation and the usage of these networks in medical fields for vessel segmentation. It further compares the various architectures based on AUC and F1 score.