Enhancing fundus image analysis for diabetic retinopathy using CheXNet with CBAM and Grad-CAM visualization

The results indicate that incorporating CBAM improves discriminative feature learning within a DenseNet121-based framework. While the model demonstrates strong performance across two public datasets, further prospective evaluation and external validation are required to assess its clinical applicability in real-world settings.