EYE DEEP- NET A MULTI CLASS DIAGNOSIS OF RETINAL DISEASES USING DEEP NEURAL NETWORK

Abstract

Retinal diseases are major causes of vision impairment and blindness, making early and accurate diagnosis essential for effective treatment. Manual examination of retinal fundus images is often time-consuming, subjective, and dependent on experienced ophthalmologists. This work proposes EyeDeep-Net, an automated deep neural network framework for multi-class diagnosis of retinal diseases from color fundus images. The proposed system performs image resizing, normalization, contrast enhancement using CLAHE, noise reduction, and data augmentation to improve image quality and model generalization. A convolutional neural network automatically extracts hierarchical retinal features, including blood vessels, optic-disc structures, hemorrhages, exudates, and other pathological abnormalities. Batch normalization, ReLU activation, max pooling, and Adam optimization are incorporated to improve training stability and performance. The Softmax classifier categorizes images into Diabetic Retinopathy, Glaucoma, Age-related Macular Degeneration, Hypertensive Retinopathy, Cataract, or Normal Retina. Performance is evaluated using accuracy, precision, recall, F1-score, sensitivity, specificity, and AUC-ROC for reliable computeraided retinal screening. Keywords: Retinal Diseases, EyeDeep-Net, Deep Learning, CNN, Fundus Images, Medical Diagnosis.