Abstract

Disease screening is critical for early detection and timely intervention in clinical practice. However, most current screening models for medical images suffer from limited interpretability and suboptimal performance, often lacking effective mechanisms to reference historical cases or provide transparent reasoning pathways. To address these challenges, we introduce EviScreen, an evidential reasoning framework for disease screening that leverages region-level evidence from historical cases. EviScreen offers retrospection interpretability through regional evidence retrieved from dual knowledge banks. With this evidential mechanism, the evidence-aware reasoning module makes predictions using both the current case and retrieved evidence from historical cases, improving screening performance while providing a more transparent reasoning process. It also enhances localization interpretability by leveraging abnormality maps derived from contrastive retrieval rather than relying on post-hoc saliency maps. Experiments on real-world disease screening benchmarks show superior performance, including higher specificity at clinical-level recall.