ABSTRACT
Diabetes is a condition that occurs when the pancreas, an important organ, does not secrete sufficient insulin, which slowly affects the retina, finally leads to a medical condition called Diabetic Retinopathy (DR). The two symptoms visible are appearance of exudates and appearance of hemorrhages. The basic difference between an exudates and hemorrhages are that exudates are secreted by the glands and hemorrhages takes place in the retina. The severity of disease is proved by the appearance and occurrence of these features. Thus, we implemented an automated computer aided system, which accurately detects these symptoms from the retinal fundus images. We have employed convolutional neural network (CNN) for detection of the various symptoms in the fundus images like exudates and hemorrhage. The CNN-based technique has been preferred due to its high statistical efficiency & computational efficiency. The experimental results on standard fundus images dataset validates, that in comparison to standard machine learning (ML) methods, the deep CNN-based method has considerably improved the classification accuracy.
