Retinal Fundus Image Analysis and Classification for Diabetic Retinopathy

Diabetic retinopathy, a condition affecting individuals with diabetes, damages the retina and can lead to vision impairment over time. Traditional screening for diabetic retinopathy is a labor-intensive process conducted by ophthalmologists. The proposed method involves training a pre-existing model on a vast dataset comprising approximately 3662 training images. The model will initially analyze the fundus eye images, extracting relevant features. Subsequently, the activation function will generate the output, determining the severity level of the patient's condition based on calculated weights. By leveraging deep learning to analyze high-resolution fundus images, the labor-intensive process of diabetic retinopathy screening can be streamlined. The deep learning model, extracts critical features from the images which enable accurate predictions of the different stages of diabetic retinopathy, potentially improving early diagnosis and patient outcomes. The proposed model attained an accuracy of 81.7% and 84.4% for VGG16 and RESNET50 respectively for diabetic retinopathy diagnosis.