Diabetic Retinopathy Detection in Retinal Fundus Images Using Sequential Convolutional Neural Network

Diabetic Retinopathy (DR) is a dynamic, incurable microvascular condition of diabetes mellitus that produces floaters, blurred vision, trouble seeing colors, and occasionally permanent blindness. Therefore, systematic discussion and review of related approaches and findings are necessary for DR detection. The analysis and identification of various ophthalmologic conditions, including arteriosclerosis, glaucoma, hypertension, DR, and choroidal neovascularization, are the main uses of fundus images of the retina. By examining retinal fundus images, the Computer-Aided Detection approach helps ophthalmologists identify such anomalies. Deep Learning is a developing and highly promising method that is used by machines to learn features of certain objects and recognize patterns, and due to the increase in the amount of data collected in the medical field, it is becoming a great option to be implemented in the field. Thus, this study will provide a thorough analysis of the issues related to retinal vessel segmentation, evaluating various extraction techniques and providing a brief example of quantitative performance metrics for retinal vessel segmentation, such as sensitivity, specificity, and accuracy, among others. Sequential Convolutional Neural Network is employed here for the classification of the APTOS fundus image dataset for the detection of Diabetic Retinopathy.