Retinal Image Analysis to Detect Neovascularization using Deep Segmentation

The retina has a significant role in early detection of sight-threatening disease symptoms. Most of the ocular complications manifest themselves in retina. The extraction of useful information from this vital resource is a critical task. The recent advancement in artificial intelligence has opened ways to provide rapid assistance in detecting ocular disorders through retinal images. In this article, we have proposed a vessels segmentation model for the early detection of neovascularization. It is a common symptom for patients facing chronic diabetic retinopathy. In neovascularization, the tiny vessels are produced that gets block over time with an extensive amount of sugar content in human blood. The detection of newly formatted tiny blood vessels needs a precise vessels extraction system. Our model has shown promising results on a publicly available retinal image dataset. It has achieved the highest accuracy of 0.9554 with 0.9780 AUC. The underlying research is an effort to produce automated disease detection system. The core function of the proposed system is to analyze the structural variation in vessels of subjects experiencing ocular disease symptoms and to reduce the risk of blindness through early diagnosis.