Back to all papers

Enhancing radiology workflows through collaborative AI-assisted chest X-ray reporting using large vision-language models: a proof-of-concept study.

April 28, 2026pubmed logopapers

Authors

Pellegrini C,Özsoy E,Gassert FT,Marka AW,Strenzke M,Keicher M,Makowski MR,Navab N

Affiliations (5)

  • School of Computation, Information and Technology, Technical University of Munich, Munich, Germany. [email protected].
  • Munich Center of Machine Learning, Technical University of Munich, Munich, Germany. [email protected].
  • School of Computation, Information and Technology, Technical University of Munich, Munich, Germany.
  • Munich Center of Machine Learning, Technical University of Munich, Munich, Germany.
  • Institute for Diagnostic and Interventional Radiology, School of Medicine and Health, TUM Klinikum, Technical University of Munich (TUM), Munich, Germany.

Abstract

To evaluate whether collaborative assistance from an artificial intelligence-based tool that proposes partial radiology report content can improve reporting efficiency and radiologist satisfaction in chest X-ray interpretation, without compromising report quality. In a retrospective study, three radiologists reported 50 MIMIC-CXR chest X-rays twice, once with artificial intelligence (AI) assistance and once without. A specialized large vision-language model (LVLM) provided real-time suggestions, which could be accepted, modified or rejected. The study evaluated writing time, suggestion acceptance, report length and quality and assessed usability and suggestion quality on a 5-point Likert-scale questionnaire. Statistical analysis used paired t-tests or Wilcoxon signed-rank tests based on normality. AI assistance reduced mean writing time by 7.80% (p = 0.08), with significant gains for complex reports (18.34%, p < 0.001). Efficiency improvements correlated with suggestion acceptance and were user-dependent, with benefits up to 27.24% (CI: [17.34, 37.14], p < 0.001) for radiologists with high acceptance. Report quality and length remained stable, indicating preserved diagnostic accuracy without degradation. Radiologists rated the tool highly for ease of use (mean: 4.33) and desired regular use (mean: 4), noting minimal errors (mean: 1.67). Collaborative AI assistance with an LVLM can improve reporting efficiency if well adopted, particularly for complex cases, without compromising quality, and is well-received by radiologists. These exploratory findings suggest potential to optimize radiology workflows through collaborative reporting and warrant prospective validation in clinical settings. This study critically evaluates a collaborative AI-assisted reporting tool for chest X-rays, demonstrating its potential to enhance radiologist efficiency without compromising automatically measured report quality, thereby demonstrating a potential path for practical integration of AI into clinical radiology workflows. A collaborative vision-language model supported radiology workflow is proposed, and its effectiveness is studied in a user study. Mean writing time for a radiology report decreases with AI support without affecting report quality. The AI-assisted tool was rated highly for usability and integration into clinical workflow, supporting its practical adoption in radiology reporting.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.