Decoding the Eye: AI-Driven Diabetic Retinopathy Classification for Precision Healthcare

Diabetic Retinopathy (DR) stands as a predominant cause of blindness in individuals with diabetes, demanding laborious manual screening processes for accurate diagnosis. DR develops as a consequence of an extended duration of diabetes. The consequential delay in this procedure poses a significant obstacle to timely intervention, emphasizing the imperative for an automated and precise approach to expedite the identification and classification of DR from retinal images. DR has been diagnosed using a variety of methods in the past, such as fuzzy approaches, image processing techniques, and machine learning algorithms like Support Vector Machine (SVM). This research harnesses the capabilities of Convolutional Neural Networks (CNNs) to craft an advanced model specially designed for the nuanced categorization of DR severity. This work examined different pre-trained models for their effectiveness in DR classification. Each pre-trained model underwent assessment using accuracy, confusion matrix, and Cohen's kappa score. Notably, the DenseNet121 model emerged as the frontrunner, showcasing exceptional performance.