A Smartphone Based Application for Early Detection of Diabetic Retinopathy Using Normal Eye Extraction

摘要

The leading cause for the blindness in the diabetic patients is due to the Diabetic Retinopathy. Diabetes is the key causes of retinal blindness for millions of people. Current detection of diabetes from retinal images using Fundus camera is expensive and inconvenient since such detection is not portable and requires specialists to perform an operation. This paper presents an innovative development of a low cost smart phone based application that allows patients in remote and isolated areas for regular eye examinations and disease diagnosis. This mobile diagnosis system uses a KNN algorithm to analyze the eye images captured by the mobile phones to identify retinal disease conditions. The algorithm is first of all trained with infected and normal eye images using a mobile device and then further developed into a mobile-based diagnosis application for Android environments. The application is optimized by using the rooted method in order to increase battery lifetime and processing capacity. The proposed mobile-based application is tested and verified using by collecting the various diabetic patients from hospitals to demonstrate its merits and capabilities. There are many features present in the retina which can act as primary signs of diabetic which can be prevented by this process. The evaluation results indicate that the system shows competitive early retinal disease of diabetic detection accuracy rates and also offers in early stage detection of retinal diseases.