EI-ViT-Net: a Vision Transformer Approach for Diabetic Retinopathy's Decision Support Pipeline

One of the main causes of blindness is diabetic retinopathy (DR), which requires early identification for efficient treatment. This study presents a decision support pipeline that incorporates a gray-scaling process for enhancement of the retinal image and a Vision Transformer for classification. Using a dataset of retinal images classified as both DR and non-DR, our proposed technique achieves notable gains in performance. By using grayscaling image enhancement, the average loss was decreased from 0.31 to 0.20 and the model's accuracy rose from 85% to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 9 \%}$</tex>. These outcomes demonstrate how well gray-scaling with Gaussian smoothing and sharpening filters works to improve feature extraction and classification precision. Considering the obstacles such as computational intricacy and unpredictability in datasets, our methodology offers a resilient resolution for prompt DR identification. This might enhance patient results by means of prompt action.