Retinopathy is a severe cause of blindness. Thus, early and accurate diagnosis may avoid bad results. With this aim, for the work presented here, a novel deep neural architecture is developed for the retinopathy screening process. Optimized SCSA Feature Selection and Deep Recurrent Neural Network Weight Optimization to Improve the Accuracy and Robustness. The framework effectively processes the retinal image datasets to identify pathological patterns with the exploration and exploitation capabilities of SCSA. Experimental results show better performance compared to traditional approaches, especially concerning precision, recall, and computation efficiency. The proposed approach would be a scalable and reliable automated retinopathy diagnosis solution to help achieve improved healthcare outcomes and early intervention strategies.