Advancing Ophthalmology through Transfer Learning and Channel-Wise Attention for Retinal Disease Classification

The global incidence of ocular disorders, which impacts more than 2.2 billion people, requires innovative approaches to ensure early and accurate diagnosis. This paper explores the integration of artificial intelligence (AI) in ophthal-mology, focusing on retinal diseases like cataracts, glaucoma, and diabetic retinopathy. The utilization of AI enables remote examinations and consultations, providing a crucial advantage in early disease detection. This study introduces a comparative analysis of state-of-the-art models, combining pre-trained convolutional neural networks (CNNs) and incorporating a channel-wise attention mechanism for enhanced discriminative ability. The proposed hybrid model, combining EfficientNetBO and Inception V3 with channel-wise attention mechanism, exhibits a remarkable accuracy of 98.13%, outperforming standalone models and other combinations. The findings underscore the efficacy of combined architectures and the integration of attention mechanisms in advancing the classification of retinal diseases.