Diabetic Retinopathy State Diagnosis based on Deep Elman Neural Network

One of the main problems in medical science is the absence of recognition, detection and prediction of different illness for working precisely and practical. Diabetes is one of the illness which caught today’s people due to disregard environmental issues. Diabetes have several states which one of them is retinopathy that losing eyesight is one the most important risk of it. Microaneurysm are red spots which are the primary symptoms occurred in retina. The most important problem is early detection of retinopathy for protecting eyesight and consequences of this disease for some time delay. This research proposed a new approach for diabetic retinopathy recognition and detection by using hybrid methods. This approach have four level. In first level, pre-processing done for probabilistic noise removal and standardization of input dataset. Then Elman Neural Network (ENN) done for image segmentation based on edge detection. In the following step, dimension reduction, feature selection and extraction done by percolation theory which features are blood vessels edges and intensity of edges. At the end, combination method of SNN and percolation theory done for detection the area of retinopathy. Results show that proposed method have the better accuracy in comparison to others.

کلیدواژه ها:

Retinopathy ، Image Segmentation ، Edge Detection ، Elman Neural Network (ENN) ، Percolation Theory