ChexFract: From General to Specialized -- Enhancing Fracture Description Generation

TL;DR

This paper introduces specialized vision-language models for fracture description in chest X-ray images, significantly improving accuracy over general models and aiding in the reporting of rare but critical pathologies.

Contribution

The work develops and trains fracture-specific models using encoders from MAIRA-2 and CheXagent, enhancing fracture report generation accuracy compared to general-purpose models.

Findings

Significant improvement in fracture description accuracy

Distinct strengths and limitations identified by fracture type and location

Public release of the best-performing fracture-reporting model

Abstract

Generating accurate and clinically meaningful radiology reports from chest X-ray images remains a significant challenge in medical AI. While recent vision-language models achieve strong results in general radiology report generation, they often fail to adequately describe rare but clinically important pathologies like fractures. This work addresses this gap by developing specialized models for fracture pathology detection and description. We train fracture-specific vision-language models with encoders from MAIRA-2 and CheXagent, demonstrating significant improvements over general-purpose models in generating accurate fracture descriptions. Analysis of model outputs by fracture type, location, and age reveals distinct strengths and limitations of current vision-language model architectures. We publicly release our best-performing fracture-reporting model, facilitating future research in accurate reporting of rare pathologies.