Retinal vessel classification to predict diabetic retinopathy and retinitis pigmentosa using a deep learning pipeline

• Main Authors: Aravindan, C., Felix, A. John, Gayathri, V., Kiran, Kishor, Venkat, G. Y. Varun
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Blindness; Blood vessels; Deep learning; Diabetes; Diabetic retinopathy; Disease; Electromagnetic absorption; Eye (anatomy); Image classification; Machine learning; Medical imaging; Retina; Signs and symptoms; Vision
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0072436

In India the age group of 20+population, it is estimated that 65 million people are diagnosed with DR and retinitis pigmentosa in 2016. On assumption that around a quarter of these people have any related diseases like Diabetic Retinopathy or the retinitis pigmentosa, in these people a very minor proportion are subjected to the factor of eyesight loss or are already suffering from it. Diabetic Retinopathy is generally the sub disease of diabetes which leads to problems in the eyes. This is mainly because of the harm caused to the light absorbing part called the retina behind the eye, the blood vessels present there get damaged due to DR which further leads to these problems. In the initial stages, diabetic retinopathy has zero symptoms or the affect individuals may experience very minor vision problems. As it develops in the retina, it gradually causes entire vision loss as worsens; the development of this problem is usually noticed in people who are affected with either type one of diabetes of even noted in the later stages of diabetes. The development and impact of this disease is directly proportional to the longer the you have diabetes and the reduced control you have over your sugar levels. Retinitis pigmentosa is an inherited degenerative disease. It gradually affects the retinal part of the eye and causes night blindness and loss of side vision. In this project, we will be effectively classifying the retinal vessel in an eye based on the scanned images of the eye using deep learning algorithms to effectively identify the occurrence of diabetic retinopathy in a patient. A dataset consisting of the scanned images of the eye is trained using the resnet deep learning algorithm. Thus, this project gives an innovative solution to effectively identify the occurrence of diabetic retinopathy & retinitis pigmentosa in a patient based on the classification of the retinal vessels.