Learning Robust Representation for Joint Grading of Ophthalmic Diseases via Adaptive Curriculum and Feature Disentanglement

Diabetic retinopathy (DR) and diabetic macular edema (DME) are leading causes of permanent blindness worldwide. Designing an automatic grading system with good generalization ability for DR DME is vital in clinical practice. However, prior works either grade or independently, without considering internal correlations between them, them jointly by shared feature representation, yet ignoring potential issues caused difficult samples data bias. Aiming to address these problems, we propose a framework joint the dynamic difficulty-aware weighted loss (DAW) dual-stream disentangled learning architecture (DETACH). Inspired curriculum learning, DAW learns from simple dynamically via measuring difficulty adaptively. DETACH separates features tasks avoid emphasis on With addition DETACH, model robust representations explore achieve better performance. Experiments three benchmarks show effectiveness robustness our under both intra-dataset cross-dataset tests.