A Hybrid Wavelet-Resnet Diffusion Based Deep Learning Model for Ocular Disease Classification

Deep learning models have become indispensable in medical image analysis, particularly for detecting and classifying ocular diseases. Automated diagnostic systems are promising to improve accuracy and accessibility, especially in regions with limited access to ophthalmologists. This study introduces a novel hybrid deep learning framework that integrates wavelet transformation with ResNet50 to enhance the multi-class classification of ocular diseases. Wavelet transformation extracts fine-grained, low-frequency details from retinal images, enabling the detection of subtle, highcontrast features that traditional methods often overlook. Concurrently, ResNet50, renowned for its deep residual learning capabilities, utilizes pre-trained features to capture high-level semantic representations. Combining these techniques, the proposed model significantly improved classification performance. The hybrid model is specifically designed to enhance the accuracy and robustness of ocular disease classification, providing a reliable tool for the automated diagnosis of Glaucoma, Cataract, Age-Related Macular Degeneration (AMD), Hypertension, and Myopia. Comparative analysis shows that the hybrid model outperforms the baseline ResNet50 regarding classification accuracy and consistency across all disease classes. While the ResNet50 model achieved an overall accuracy of 94 %, the proposed hybrid model improved performance to $\mathbf{9 8 \%}$. Furthermore, F1-scores for the hybrid model reached 0.97 for Glaucoma, 0.97 for Cataract, 0.97 for AMD, 1.00 for Hypertension, and 0.98 for Myopia, demonstrating more balanced and precise predictions than the baseline. These findings underscore the hybrid model's potential for clinical applications, where fast, accurate, and automated diagnostic tools are essential for early intervention, improved patient outcomes, and broader accessibility to ocular disease screening.