Classification of diabetic retinopathy using artificial neural network

Abstract

Diabetic retinopathy (DR) is one of the complications caused by diabetes which shows abnormalities symptoms in retinopathy and is a major cause of loss of vision. The screening by an ophthalmologist is the only way to prevent this problem. This work, aim to develop classification of diabetic retinopathy algorithm by using Artificial Neural Network (ANN) for work together with telemedicine project in Thailand. First, using mathematic morphology and image processing techniques to extract features that are factor of DR. Then, input into ANN to grading the symptoms of DR. When comparing the performance of proposed software with diagnosis of ophthalmologist found that, its diagnosis have accuracy of 98.89%, sensitivity of 99.26%, specificity of 97.77% and positive predictive values at 99.26%. Thus, proposed software can helps to increase occasion of screening diabetes patients, especially in remote area where lack of ophthalmologists or specialist to read fundus images. Suitable for telemedicine system. Moreover, also can improve accuracy of ophthalmologists’s diagnosis.