DEEPDIABETIC: AN AI-POWERED DIAGNOSTIC MODEL FOR EARLY DETECTION OF RETINAL ABNORMALITIES IN DIABETIC PATIENTS

Abstract

Diabetic eye diseases, particularly diabetic retinopathy and macular edema, are among the leading causes of preventable blindness worldwide. Early detection and timely diagnosis are crucial to mitigating vision loss and improving patient outcomes. This paper presents DeepDiabetic, an AI-powered diagnostic model that leverages deep neural networks (DNNs) to automatically identify and classify retinal abnormalities from fundus images. The system aims to support ophthalmologists in clinical decision-making by providing accurate, consistent, and rapid screening of diabetic eye conditions. The proposed DeepDiabetic model utilizes a convolutional neural network (CNN) architecture trained on a large-scale dataset of retinal images preprocessed through techniques such as image normalization, contrast enhancement, and noise reduction. Feature extraction layers capture minute retinal patterns, including microaneurysms, hemorrhages, and exudates, which are critical indicators of diabetic damage. Transfer learning and fine-tuning methods are incorporated to improve performance across diverse imaging conditions and patient demographics. The model’s classification accuracy, sensitivity, and specificity are evaluated using benchmark datasets and compared with existing diagnostic frameworks. Experimental results demonstrate that DeepDiabetic achieves superior diagnostic accuracy and robustness, outperforming traditional machine learning methods and some state-of-the-art deep learning approaches. The model’s interpretability is enhanced through visualization techniques such as GradCAM, which highlight the most affected retinal regions. Furthermore, the system’s scalability enables integration into teleophthalmology platforms, allowing remote and real-time diabetic eye screening in resource-limited healthcare environments. The DeepDiabetic framework establishes a promising step toward automated, AI-assisted ophthalmic diagnosis, contributing to early disease detection, reduced screening costs, and improved accessibility to quality eye care.