Harnessing EHRs for Diffusion-based Anomaly Detection on Chest X-rays

TL;DR

This paper introduces Diff3M, a multi-modal diffusion framework that combines chest X-rays and EHR data with cross-attention to improve unsupervised anomaly detection in medical imaging.

Contribution

The paper presents a novel image-EHR cross-attention module and static masking strategy, enhancing diffusion-based anomaly detection by integrating clinical context.

Findings

Diff3M outperforms existing UAD methods on CheXpert and MIMIC-CXR/IV datasets.

Incorporating EHR data improves differentiation between normal and abnormal images.

The proposed methods achieve state-of-the-art results in medical anomaly detection.

Abstract

Unsupervised anomaly detection (UAD) in medical imaging is crucial for identifying pathological abnormalities without requiring extensive labeled data. However, existing diffusion-based UAD models rely solely on imaging features, limiting their ability to distinguish between normal anatomical variations and pathological anomalies. To address this, we propose Diff3M, a multi-modal diffusion-based framework that integrates chest X-rays and structured Electronic Health Records (EHRs) for enhanced anomaly detection. Specifically, we introduce a novel image-EHR cross-attention module to incorporate structured clinical context into the image generation process, improving the model's ability to differentiate normal from abnormal features. Additionally, we develop a static masking strategy to enhance the reconstruction of normal-like images from anomalies. Extensive evaluations on CheXpert and MIMIC-CXR/IV demonstrate that Diff3M achieves state-of-the-art performance, outperforming existing UAD methods in medical imaging. Our code is available at this http URL https://github.com/nth221/Diff3M