CFR-Net:Collaborative Feature Refnement Network for Medical Image Anomaly Detection

2026-07-13

AI Summary

This work addresses the limited adaptability of pretrained natural-image models in medical anomaly detection, which stems from fine-grained local distribution shifts, multi-scale contextual mismatches, and orientation-sensitive structural biases. To overcome these challenges, we propose the Collaborative Feature Refinement Network (CFR-Net), which leverages a frozen teacher–trainable student architecture. CFR-Net introduces a shared-parameter Multi-Path Feature Refinement Module (MPFRM) before decoding to effectively fuse general-purpose priors with domain-specific medical representations, and enforces cross-spatial consistency constraints alongside layer-adaptive learning after decoding. Combined with a variance-sensitive objective function and a dynamic “homework set” reorganization strategy, CFR-Net achieves state-of-the-art performance in both anomaly classification and precise localization across multiple medical imaging benchmarks, using only normal training data.

Abstract

Medical image anomaly detection remains challenging because networks pretrained on natural images often exhibit limited adaptability to medical images, where abnormal patterns appear as fine-grained local shifts, multi-scale contextual mismatches, and orientation-sensitive structural deviations. To address this, we propose the Collaborative Feature Refinement Network (CFR-Net), which combines shared teacher-student feature refinement before decoding with cross-space consistency after decoding. CFR-Net refines frozen teacher features and trainable student features using a Multi-Path Feature Refinement Module (MPFRM) with shared parameters, imposing common multi-path refinement rules on generic visual references and representations adapted to the medical domain, thereby mitigating domain discrepancy while modeling local, multi-scale, and orientation-sensitive feature characteristics. A variance-sensitive objective and dynamic ``homework set'' reorganization further support layer-adaptive consistency learning. Experiments on medical benchmarks show that CFR-Net achieves competitive anomaly classification and strong anomaly localization performance when trained on normal data.

Problem

• medical image anomaly detection
• domain discrepancy
• fine-grained local shifts
• multi-scale contextual mismatches
• orientation-sensitive structural deviations

Innovation

• Collaborative Feature Refinement
• Multi-Path Feature Refinement Module
• Cross-Space Consistency
• Domain Adaptation
• Anomaly Localization