A Novel Backbone for Enhanced Diabetes Retinopathy Classification

Our proposal presents a backbone for the classification of diabetes retinopathy that enhances its multi-scale representation ability. This is achieved by constructing hierarchical residual-like connections within each single radix block. Additionally, our model improves the learning of low-level feature information through an attention force mechanism that suppresses non-lesion feature information and enhances the learning of typical lesion features. As a result, the classification performance of the model is further improved. The experimental results demonstrate that our proposed method significantly increases the accuracy of DR model classification.