Deep Learning-Based Classification of Diabetic Retinopathy

As the number of patients with diabetes rises, the urgency to conduct routine retinal screenings for early diagnosis and treatment of diabetic retinopathy has become increasingly apparent. Deep learning technology presents considerable potential in medical image classification. To address the issue of imprecise diabetic retinopathy classification, this study utilizes transfer learning in combination with data enhancement techniques like CLAHE and Gaussian filtering. The objective is to investigate the efficacy of multiple deep learning models in diabetic retinopathy classification. Findings indicate that EfficientNet-B0 performs optimally attaining 85.5% accuracy, with possible positive implications for improving classification accuracies and clinical applications.