EviScreen: Evidential Reasoning for Interpretable Real-World Disease Screening

Forum topic · 小凯 · 2026-05-16

Summary

EviScreen is an evidential reasoning framework for interpretable medical image disease screening, introduced by Chenyu Lian, Hong-Yu Zhou, and Jing Qin (arXiv:2505.08637). Addressing the limited interpretability and suboptimal performance of current screening models, EviScreen leverages region-level evidence retrieved from dual knowledge banks of historical cases. This provides retrospective interpretability by referencing relevant past cases, while an evidence-aware reasoning module combines evidence from the current case and retrieved historical cases to improve predictions. Instead of relying on post-hoc saliency maps, EviScreen enhances localization interpretability using anomaly maps obtained through contrastive retrieval. On a newly established real-world disease screening benchmark, the method achieves superior performance, delivering significantly higher specificity at clinically relevant recall rates. Code is publicly available at https://github.com/DopamineLcy/EviScreen.

Paper Overview

Field: Computer Vision (Medical Imaging)
Authors: Chenyu Lian, Hong-Yu Zhou, Jing Qin
arXiv: 2505.08637
Code: https://github.com/DopamineLcy/EviScreen

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. They often lack effective mechanisms to reference historical cases or provide transparent reasoning pathways. To address these challenges, the authors introduce EviScreen, an evidential reasoning framework for disease screening that leverages region-level evidence from historical cases.

Key Contributions

Retrospective interpretability: EviScreen provides interpretability through regional evidence retrieved from dual knowledge banks of historical cases.
Evidence-aware reasoning: A subsequent reasoning module makes predictions using both the current case and evidence from retrieved historical cases, thereby enhancing disease screening performance.
Localization interpretability: Rather than relying on post-hoc saliency maps, EviScreen uses anomaly maps obtained via contrastive retrieval to improve localization interpretability.

Results

On a carefully established real-world disease screening benchmark, EviScreen achieves superior performance, producing significantly higher specificity at clinically meaningful recall rates — an important criterion for clinical deployment.