Benchmarking Robustness of Contrastive Learning Models for Medical Image-Report Retrieval

TL;DR

This paper benchmarks the robustness of four contrastive learning models for medical image-report retrieval, revealing high sensitivity to data out-of-distribution and emphasizing the need for domain-specific training to improve reliability.

Contribution

It introduces an occlusion-based robustness benchmark for contrastive models in medical retrieval and compares their performance, highlighting the importance of domain-specific training data.

Findings

All models are highly sensitive to image occlusion.

MedCLIP shows slightly better robustness than others.

General-purpose trained CLIP performs poorly on medical data.

Abstract

Medical images and reports offer invaluable insights into patient health. The heterogeneity and complexity of these data hinder effective analysis. To bridge this gap, we investigate contrastive learning models for cross-domain retrieval, which associates medical images with their corresponding clinical reports. This study benchmarks the robustness of four state-of-the-art contrastive learning models: CLIP, CXR-RePaiR, MedCLIP, and CXR-CLIP. We introduce an occlusion retrieval task to evaluate model performance under varying levels of image corruption. Our findings reveal that all evaluated models are highly sensitive to out-of-distribution data, as evidenced by the proportional decrease in performance with increasing occlusion levels. While MedCLIP exhibits slightly more robustness, its overall performance remains significantly behind CXR-CLIP and CXR-RePaiR. CLIP, trained on a general-purpose dataset, struggles with medical image-report retrieval, highlighting the importance of domain-specific training data. The evaluation of this work suggests that more effort needs to be spent on improving the robustness of these models. By addressing these limitations, we can develop more reliable cross-domain retrieval models for medical applications.