Diabetic retinopathy diagnosis using neural network arbitration

Abstract

In this research work, we have implemented an intelligent system for the diagnosis of diabetic retinopathy. Diabetic retinopathy is damage caused by diabetes to the retinal blood vessels. This causes the leak of blood and other fluid that resulted in swelling of retina tissue and cloudy vision. This can be classiced into two; non-proliferative diabetic retinopathy and proliferative diabetic retinopathy. The dataset used for the implementation of this intelligent system is obtained from the freely available UCI machine learning repository. This system will aid physicians in accurately diagnosing the disease. Such a system will make the diagnosis faster, more accurate, and easier as compared to manual diagnosis. Our novel system uses a feedforward neural network trained with a backpropagation neural network. The results obtained in this work are compared with previously proposed systems using the same dataset. Experimental results indicate that our novel system outperforms the other systems in diagnosing diabetic retinopathy.