Generation, Analysis and Validation of Retinal Images Associated with Diabetic Retinopathy Using Generative Artificial Intelligence

Diabetic retinopathy is a major cause of vision loss globally, and its early detection contributes to preventing complications. In this context, an architectural framework is proposed for retinal image generation in diabetic retinopathy. The generated images are subjected to a systematic clinical validation process conducted by ophthalmologists to assess their alignment with the clinical manifestations of diabetic retinopathy and enhance their applicability in pathology analysis. The dataset used for training, testing, and validating the generative model must possess an adequate volume and a balanced class distribution. High visual quality and accurate representation of the morphological features specific to each stage of diabetic retinopathy are fundamental elements for achieving the targeted outcomes. The developed architecture enables the synthesis of highly realistic retinal representations, supporting both the training of diagnostic models and the investigation of disease progression through generative artificial intelligence.