A two-stage deep learning framework with progressive feature fusion and attention mechanisms for multi-class retinal disease classification
Results show that the proposed framework provides a good trade-off between classification performance, computational requirements and interpretability and shows potential for automated OCT-based retinal disease screening support.