Accurate alignment between chest X-ray (CXR) images and radiology reports is essential for clinical safety, yet automatic radiology report error detection from paired CXRs remains challenging. Prior approaches typically compare the entire report with the image, limiting the model’s ability to learn fine-grained vision-language correspondences at the level of individual findings. We reformulate report error detection as a sentence-level vision-language learning problem by decomposing reports into atomic finding statements, thereby reducing semantic entanglement and supervision noise. We propose SenRED (Sentence-level Report Error Detector), which uses sentence-conditioned patch aggregation to associate each statement with localized visual evidence. To supervise learning, we generate sentence-level synthetic errors reflecting clinically relevant categories—false positives, false negatives, and mislocalization—along with consistent image-sentence pairs. Experiments on MIMIC-CXR and Open-i show that SenRED consistently outperforms task-specific baselines and zero-shot instruction-tuned VLMs in both sentence-level and exam-level evaluations. Radiologist-confirmed evaluations further show that these gains are maintained on clinically reviewed subsets. Comparative experiments support the benefit of sentence-level reformulation, while ablations demonstrate that sentence-conditioned patch aggregation improves fine-grained consistency checking.

Radiology report is an official record of radiologists’ interpretation of patients’ radiographs and it’s a crucial component in the overall medical diagnostic process. However, it can contain various types of errors that can lead to inadequate treatment or delay in diagnosis. To address this problem, we propose a deep learning framework to detect errors in radiology reports. Specifically, our method detects errors between findings and conclusion of chest X-ray reports based on a supervised learning framework. To compensate for the lack of data availability of radiology reports with errors, we develop an error generator to systematically create artificial errors in existing reports. In addition, we introduce a Medical Knowledge-enhancing Pre-training to further utilize the knowledge of abbreviations and key phrases frequently used in the medical domain. We believe that this is the first work to propose a deep learning framework for detecting errors in radiology reports based on a rich contextual and medical understanding. Validation on our radiologist-synthesized dataset, based on MIMIC-CXR, shows 0.80 and 0.95 of the area under precision-recall curve (AUPRC) and the area under the ROC curve (AUROC) respectively, indicating that our framework can effectively detect errors in the real-world radiology reports.