Automated Detection of Diabetic Retinopathy Using Vision Transformers: A Deep Learning Approach

Diabetic Retinopathy (DR) is a prevalently occurring impediment of diabetes mellitus. It is also considered to be one of the leading causes of visual impairment in the world. The importance of timely intervention and treatment is vital in finding the stages of DR and classifying them accordingly in the initial stages of detection. This study enhances a deep learning resolution on Vision Transformers (ViT) in image automated detection and grading of DR on retinal fundi images. The proposed model processes high-resolution retinal images by dividing them into patches before applying the self-attention mechanisms to identify more intricate features that indicate the extent of the DR severity by relying on the APTOS 2019 Blindness Detection dataset. The measures of the model performance are accuracy, precision, recall, and F1-score which demonstrate the efficiency of the model in classifying the different stages of DR. The results indicate the potential of ViT-based architectures in enhancing the procedure of screening offered by DR, which is a promising method in ophthalmic diagnosis.