Diabetic Retinopathy Identification Using Optimized Deep CNN Classifier

Diabetic Retinopathy (DR) represents general complications arising from diabetes, primarily resulting from elevated blood glucose levels. This research project is primarily focused on the design and development of a precise detection model for DR. The research workflow begins with the acquisition of retinal fundus images, followed by a sequence of preprocessing steps and Region of Interest (ROI) extraction to enhance critical image features essential for subsequent analysis. This research is characterized by its integration of critical elements, beginning with the segmentation of retinal blood vessels. Additionally, it introduces an improved Generative Adversarial Network (GAN) tailored for more efficient image semantic segmentation. Following the segmentation process, the research utilizes feature extraction techniques, including Hybridized Ternary Patterns and the GrayLevel Co-occurrence Matrix (GLCM). The Deep Convolutional Neural Network (DCNN), classifies the extracted feature based on the distinctive attributes of the input image, which is fine-tuned by the Coyote optimizer (COA). The proposed COA-DCNN achieves an accuracy of 95.05%, and 94.21% for training percentage (TP), and k-fold respectively, which underscores the efficacy and superiority of the COA-DCNN approach, highlighting its potential to make a substantial impact on the realm of medical image analysis. It offers a noteworthy contribution that could significantly improve the diagnosis of DR, holding great promise for this critical healthcare domain.