Reinforcement learning improves accuracy of chest x-ray reports
Enhancing Visual Reasoning in Chest X-Ray Report Generation Using Reinforcement Learning
Computer Vision and Pattern Recognition
Summary
Chest x-ray reports need to be both accurate and easy to understand, but current computer models often make mistakes by copying text without checking facts. The authors improved these models by teaching them to check their reasoning using pictures and descriptions of specific body parts. They used a special learning method called reinforcement learning to help the model think step-by-step and avoid errors. This approach helped make the reports more reliable and easier for doctors to check.
What this means in practice
- •For medical imaging teams: Generate chest x-ray reports with improved accuracy and clearer reasoning steps for easier clinical review and auditing.
- •For radiology software developers: Integrate structured reasoning verification into radiology AI tools to reduce factual errors in generated medical reports.
Authors
Denis Musinguzi, Andrew Katumba, Prasenjit Mitra
Abstract
Medical report generation has made significant progress with the rise of modern vision-language models and the growing availability of large-scale medical datasets. However, hallucinations remain a major challenge, largely due to the limitations of supervised fine-tuning (SFT), which prioritizes lexical similarity to reference reports rather than clinical correctness. While reinforcement learning has shown strong performance in domains with verifiable rewards such as mathematics and code generation, its application to open-ended medical tasks remains limited. Existing work focuses on evaluating final answers, overlooking the model's reasoning, despite evidence that flawed reasoning can degrade overall performance. In this study, we propose a framework that verifies the model's reasoning process by integrating anatomical regions, bounding boxes, and region-level textual descriptions. We design spatial and factual reward mechanisms to ensure that the model's reasoning is both visually grounded and factually accurate. Starting from Qwen3-VL-8B-Instruct as our base model, we adapt it to the medical domain using supervised fine-tuning, introduce reasoning capability through a cold-start SFT stage, and refine it with reinforcement learning. We find that RL provides performance gains beyond those achievable through SFT alone, and that jointly verifying both reasoning steps and final outputs yields larger improvements than verifying either in isolation. We further identify multiple modes of reward hacking in the RL stage. Finally, the model's structured think traces enhance interpretability, making its outputs easier to audit for clinical use.