Deep Learning Approaches for Detecting Diabetic Retinopathy using CNN Models

Chronic diabetes is linked to diabetic retinopathy (DR), an eye condition. Is therefore predicted will be even more than DR affects 93 million populations? Making it the main source of vision among adults of working age. Early detection of DR is essential for preventing severe vision loss; however, this may be challenging since the condition often displays no indications until it's too late to give effective therapy. Retinal fundus views in digitized RGB are now examined and evaluated by ophthalmologists or qualified physicians to identify DR through the presence of lesions associated with the disease's pathological changes, a procedure that is both time-consuming and labor-intensive. The development of DR may be controlled or managed by using the computerized method of diagnostic testing to expedite detection and consideration. The previous techniques used to Support Vector Machine and Probabilistic Neural Network techniques are poor results. This research work provides an automated classification method, which uses a convolutional neural network, VGG-16, as well as VGG-19 machine learning methods were used to analyze fundus images with varying illumination and sectors of vision and produce a seriousness grading for diabetic retinopathy (DR). Classifying pictures 80% sensitive, 82% precision, 82% specific, and a range of 0 to 4, where 0 represents no DR and 4 represents proliferative DR 0.904 AUC is an impressive feat for this system.