Reconstruction-Shift Discrimination via Mask-Guided Latent Diffusion for Medical Anomaly Detection

2026-08-01

AI Summary

This work addresses the challenge in unsupervised medical anomaly detection where ambiguous reconstruction residuals often lead to either retention of lesions or misclassification of normal anatomical variations and noise as anomalies. To overcome this, the authors propose a Discriminative Mask-guided Diffusion (DMD) framework that learns compact quantized latent representations of normal images and introduces local mask-based perturbations to latent variables, followed by latent diffusion-based reconstruction. This process forms a self-supervised binary classification task, enabling simultaneous image-level anomaly scoring and pixel-level localization. The method innovatively integrates a reconstruction-shift discrimination mechanism, fusing mask-guided perturbation with diffusion reconstruction to establish a paradigm distinct from conventional residual analysis. Evaluated across five datasets—including brain MRI, breast ultrasound, and chest X-ray—DMD consistently outperforms current state-of-the-art approaches.

Abstract

Unsupervised medical anomaly detection learns normal anatomical patterns from healthy training images and identifies deviations at test time. Reconstruction-based and diffusion-based methods commonly use the difference between an input image and its reconstruction as anomaly evidence. However, this residual can be ambiguous. Expressive models may preserve pathological structures, while benign anatomical variation, imaging noise, and acquisition differences may also produce large reconstruction errors. We propose discriminative mask-guided diffusion (DMD), a medical anomaly detection framework that complements residual-based localization with reconstruction-shift discrimination. DMD first learns a compact quantized latent representation of normal images. Localized masks then perturb selected latent regions, and a latent diffusion model reconstructs the perturbed representations. The resulting reconstructions are paired with their original normal images to define a self-supervised classification task. At inference, the classifier provides a learned image-level anomaly score, while the residual between the input and its diffusion-based reconstruction yields a pixel-level anomaly map. Experiments on five datasets spanning brain MRI, breast ultrasound, and chest radiography show that DMD achieves the best overall performance among the state-of-the-art baseline methods.

Problem

• unsupervised medical anomaly detection
• reconstruction ambiguity
• residual-based localization
• anatomical variation
• imaging noise

Innovation

• mask-guided diffusion
• anomaly detection
• latent representation
• self-supervised classification
• medical imaging