Diabetic retinopathy is one of the most common microvascular complications of diabetes mellitus and remains a leading cause of preventable vision impairment worldwide. Early detection and timely intervention are important for reducing disease progression and preserving visual function. Retinal fundus imaging is widely used for screening diabetic retinopathy; however, manual assessment requires specialized expertise and may be challenging in large-scale screening programs. Recent advances in deep learning have demonstrated promising results in medical image analysis and automated disease detection. This study proposes a hybrid deep learning model for automated detection and classification of diabetic retinopathy using retinal fundus images. The framework combines convolutional neural network-based feature extraction with a feature fusion and classification module to improve discrimination across different disease stages. Image preprocessing techniques were applied to enhance retinal structures and reduce image variability before model training. The developed model was evaluated using publicly available retinal imaging datasets containing normal and diabetic retinopathy cases with multiple severity levels. Performance was assessed using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve. Experimental results indicate that the proposed hybrid model effectively identifies retinal abnormalities associated with diabetic retinopathy and provides reliable classification across disease categories. The findings suggest that deep learning-based screening systems may support ophthalmologists in large-scale retinal assessment and facilitate earlier diagnosis. The proposed framework contributes to automated diabetic retinopathy screening by combining image-based learning with clinically relevant disease classification.