TransXplainRRG: A Radiologist-Evaluated LVLM for Chest Radiograph Report Generation
Organizations: Curtin University, Bentley, Western Australia · Fiona Stanley Hospital, Murdoch, Western Australia
Abstract
Recent advancements in Large Vision-Language Models (LVLM) enhance X-ray report generation, but often neglect clinical context, overlook workflow integration, and lack explainability, which hinders clinical adoption due to interpretability and reliability concerns. This paper addresses the dual challenges of performance and expert explainability in automated X-ray report generation for clinical AI applications. Our proposed approach, TransXplainRRG, utilises an off-the-shelf Swin Transformer model along with a transformer-based text encoder that incorporates patient medical history to generate a radiology report. Further, we explore an expert-guided `inside-out' approach and extract only abnormal findings for radiology report refinement. Thus, this study bridges the gap between high-performance automation and interpretability critical for clinical practice by combining state-of-the-art transformer-based vision encoders, text encoders, and LLMs. TransXplainRRG is trained and tested on the large-scale MIMIC-CXR dataset and further evaluated on the unseen IU X-Ray dataset to demonstrate its generalizability and robustness. We employed the GREEN metric as a complementary measure of clinical consistency, achieving scores of 0.327 on MIMIC-CXR and 0.605 on IU X-Ray, both higher than those achieved by MAIRA-2 under the same evaluation setting. It also performed competitively across standard evaluation metrics, frequently matching or surpassing state-of-the-art models. Additionally, we introduce qualitative evaluation metrics developed from radiologists' viewpoints and conduct a preliminary radiologist evaluation on 100 generated reports to assess their clinical relevance. Quantitative and qualitative analyses further demonstrate the potential of our TransXplainRRG framework for reliable and explainable radiology report generation.
Figures & tables
| Models | Test Samples | Multi-Views | B4 | METEOR | ROUGE-L | CIDER |
| MedXChat 1 | 3858 images | No | 0.111 | 0.135 | 0.264 | 0.175 |
| Nguyen et al. 2021 [ 12 ] | 1000 Studies | Yes | 0.224 | 0.222 | 0.390 | – |
| UniXGen-256 1 | 2733 Studis | Yes | 0.101 | 0.156 | 0.294 | 0.138 |
| 1 | 1624 studies | Yes | 0.083 | — | 0.269 | 0.187 |
| CXRMate 2024 [ 17 ] | 1624 studies | Yes | 0.074 | 0.158 | 0.255 | 0.172 |
| MAIRA-2 2024 [ 16 ] | 2461 studies | Yes | 0.231 | 0.417 | 0.384 | – |
| Error Category | Representative Observation |
| Missed findings | Clinically relevant masses, nodules, opacities, fractures, or lung abnormalities were omitted. |
| Support-device errors | Device presence, position, or laterality was incorrectly reported or omitted, including NGT, PICC, ETT, and chest-tube information. |
| Anatomical localisation errors | Findings were occasionally assigned to the wrong anatomical side or region. |
| False-positive findings | Findings not supported by the reference report were occasionally introduced, such as an absent port-a-cath or emphysema. |
Radiology Report Generation
Radiology report generation uses medical images and clinical context to produce structured, clinically accurate findings and impressions.
See all 170 papers