Automatic Classification of Retinal Fundus Images for Diabetic Retinopathy Detection

Retinal imaging facilitates the opinion and helps to detect diabetic retinopathy at the early stage. Cardiovascular conditions of the human body can be identified by analyzing the retinal images and the risk is tracked by analyzing blood vessels of the retina. Discovery of any abnormal features in the eye fundus retinal in the before stage is essential as well as experts on this sphere need further acceptable automatic discovery styles to describe the unheroic spots caused by diabetics to control its propagation on the retina. Image analysis of the retina is the fundamental approach to handle the eye diseases and other infections affecting eye fundus. This system introduces a more accurate medical opinion model for diabetic retinopathy discovery in fundus images by automated classification of landmarks. (i) Data preprocessing and augmentation; (ii) Automated classification with deep learning; (iii) Principal feature selection with Whale optimization Algorithm; (iv) Feature extraction; (v) Performance assessment. In the proposed project we used a class of automatic classification algorithms in our system, and the VGG-16 produces better results. The model secured an accuracy of 94 % and loss function of 0.0633.