As population of diabetic has increased worldwidely, supply for test of diabetic retinopathy has also been increasing. However, current testing method has limitations on manpower, cost, and sustainability. As an alternative, automatic testing by using deep learning system has been proposed, yet it requires validation on mutli-ethincity and environment data. Also, not only detecting DR(diabetic retinopathy), it is important to detect other diseases such as glaucoma and AMD.
The purpose of research was to evaluate to what extant can DLS(deep learning system) accurately detect Referable diabetic retinopathy, Vision-threatening diabetic retinopathy, Referable possible glaucoma, and Age-related macular degeneration (AMD). In addition, This study evaluated the feasibility of applying a deep learning system (DLS) in both a fully automated screening model and a semiautomated screening model that combines artificial intelligence with human graders.
The results were impressive. DLS showed high sensitivity and specificity for detecting referable and vision-threating diabetic retinopathy on multiethnic population groups. Moreover, DLS also showed strong performance in identification in regards to referable plausible glaucoma and age-related macular degeneration(AMD) These crucial findings suggest that DLS can assist in detecting disease in premediated way, giving hope to diabetes patients
