Revolutionizing AMD detection Bi model CNNs and hybrid feature selection for automated grading

Age-related macular degeneration (AMD) is a common cause of vision loss in older adults. The automated grading of AMD from fundus images can aid in early detection and treatment. In this research, we propose a comprehensive framework that can enhance the quality of fundus images and increase automated grading accuracy. We employ preprocessing techniques such as the Adaptive Contrast Enhancement Algorithm (CLAHE) and Gamma correction to improve image quality and feature representation. Additionally, we suggest a hybrid feature selection approach that combines handcrafted and deep learning-based features to identify discriminative features for AMD grading. This approach can comprehensively represent AMD-related abnormalities, leaving no room for ambiguity. Furthermore, we introduce a new Bi-Model Convolutional Neural Network (CNN) architecture that uses global and local features for accurate AMD grading. The Bi-Model CNN integrates feature representations extracted from the entire fundus image and local patches to capture global context and fine-grained details. This approach offers precise, consistent, and prompt results, thereby increasing the effectiveness of early disease identification. Experimental results demonstrate that the Bi-CNN + Feature Fusion model achieves a 99.5% accuracy, outperforming other configurations. The model's performance is further validated with high precision (0.995), recall (0.995), and F1-score (0.995), along with a Cohen's Kappa of 0.990, indicating almost perfect agreement between predicted and actual labels.