CorBenchX: Large-Scale Chest X-Ray Error Dataset and Vision-Language Model Benchmark for Report Error Correction

TL;DR

This paper introduces CorBenchX, a large-scale dataset and benchmark for error detection and correction in chest X-ray reports, and proposes a reinforcement learning framework to improve model performance in clinical report correction.

Contribution

It provides the first large-scale dataset for chest X-ray report error correction and benchmarks multiple vision-language models, proposing a novel reinforcement learning method to enhance correction accuracy.

Findings

o4-mini achieves 50.6% detection accuracy

MSRL improves detection precision by 38.3%

MSRL enhances correction scores by 5.2%

Abstract

AI-driven models have shown great promise in detecting errors in radiology reports, yet the field lacks a unified benchmark for rigorous evaluation of error detection and further correction. To address this gap, we introduce CorBenchX, a comprehensive suite for automated error detection and correction in chest X-ray reports, designed to advance AI-assisted quality control in clinical practice. We first synthesize a large-scale dataset of 26,326 chest X-ray error reports by injecting clinically common errors via prompting DeepSeek-R1, with each corrupted report paired with its original text, error type, and human-readable description. Leveraging this dataset, we benchmark both open- and closed-source vision-language models,(e.g., InternVL, Qwen-VL, GPT-4o, o4-mini, and Claude-3.7) for error detection and correction under zero-shot prompting. Among these models, o4-mini achieves the best performance, with 50.6 % detection accuracy and correction scores of BLEU 0.853, ROUGE 0.924, BERTScore 0.981, SembScore 0.865, and CheXbertF1 0.954, remaining below clinical-level accuracy, highlighting the challenge of precise report correction. To advance the state of the art, we propose a multi-step reinforcement learning (MSRL) framework that optimizes a multi-objective reward combining format compliance, error-type accuracy, and BLEU similarity. We apply MSRL to QwenVL2.5-7B, the top open-source model in our benchmark, achieving an improvement of 38.3% in single-error detection precision and 5.2% in single-error correction over the zero-shot baseline.