EUVIMEDEUVIMED entdecken
QuerschnittstudieOpen Access

Diagnostic Accuracy and Fine-Grained Image-Interpretation Performance of Multimodal Large Language Models on Bilingual Radiology Questions: Cross-Sectional Benchmark Study

Yijie Liu, Xin Zhang, Jia Xu, Manxin Huang, Yinting Xiong, Wenqian Wu

Journal of Medical Internet Research · 11.11.2026 · 28 · e98849-e98849

DOI: 10.2196/98849 ↗

Kurz gefasst

Querschnittstudie in der Zeitschrift Journal of Medical Internet Research, online erschienen am 11.11.2026. Verfasst von Yijie Liu und 5 weiteren Autor:innen. Beteiligte Einrichtung: Union Hospital, Department of Ultrasound Medicine, Tongji Medical College, Huazhong University of Science and Technology, 1277 Jiefang Avenue, Wuhan,…. Hauptthemen laut MEDLINE: Multilingualism, Radiology, Image Interpretation, Computer-Assisted. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Multimodal large language models (LLMs) are rapidly expanding in radiology and digital health; however, their diagnostic performance across diverse clinical scenarios and fine-grained visual reasoning tasks requires systematic evaluation.

Objective

The study aims to evaluate the diagnostic accuracy and fine-grained visual reasoning performance of 5 multimodal LLMs with bilingual medical imaging questions.

Methods

This cross-sectional benchmark study evaluated bilingual medical imaging questions using zero-shot prompting on 5 LLMs: GPT-5.2, Gemini 3 Pro, Doubao, Yuanbao, and Llama 4 Maverick. Subgroup analyses by language (English/Chinese), question type (image/text), imaging modality (computed tomography, radiography, magnetic resonance imaging, ultrasound, and nuclear medicine), and anatomical system (neurologic, respiratory, cardiovascular, digestive, genitourinary, and musculoskeletal) evaluated intermodel accuracy and intramodel consistency. Image-based visual reasoning performance was assessed across modality, anatomy, lesion detection, and localization.

Results

A total of 298 questions (198 image-based, 100 text-based) were evaluated. Gemini 3 Pro achieved the highest overall accuracy (n=268, 89.9%, 95% CI 86.0%‐92.9%), significantly outperforming Doubao (n=235, 78.9%; P

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

MultilingualismRadiologyImage Interpretation, Computer-AssistedLarge Language ModelsCross-Sectional StudiesHumansBenchmarking

Stichwörter der Autor:innen

AIbenchmarkinglarge language modelsmedical imagingmultiple-choice questionsradiology

Autor:innen und Institutionen

  • Union Hospital, Department of Ultrasound Medicine, Tongji Medical College, Huazhong University of Science and Technology, 1277 Jiefang Avenue, Wuhan, 430022, China., Wuhan, Hubei, China, 86 27-85726430, 86 27-85726430.
  • Hubei Province Clinical Research Center for Medical Imaging, Wuhan, Hubei, China.
  • Hubei Province Key Laboratory of Molecular Imaging, Wuhan, Hubei, China.

Zitieren

APA

Liu, Y., Zhang, X., Xu, J., Huang, M., Xiong, Y., & Wu, W. (2026). Diagnostic Accuracy and Fine-Grained Image-Interpretation Performance of Multimodal Large Language Models on Bilingual Radiology Questions: Cross-Sectional Benchmark Study. Journal of Medical Internet Research, 28, e98849-e98849. https://doi.org/10.2196/98849

Vancouver

Liu Y, Zhang X, Xu J, Huang M, Xiong Y, Wu W. Diagnostic Accuracy and Fine-Grained Image-Interpretation Performance of Multimodal Large Language Models on Bilingual Radiology Questions: Cross-Sectional Benchmark Study. Journal of Medical Internet Research. 2026;28:e98849-e98849. doi:10.2196/98849

BibTeX
@article{Liu2026,
  author = {Liu, Yijie and Zhang, Xin and Xu, Jia and Huang, Manxin and Xiong, Yinting and Wu, Wenqian},
  title = {Diagnostic Accuracy and Fine-Grained Image-Interpretation Performance of Multimodal Large Language Models on Bilingual Radiology Questions: Cross-Sectional Benchmark Study},
  journal = {Journal of Medical Internet Research},
  year = {2026},
  volume = {28},
  pages = {e98849--e98849},
  doi = {10.2196/98849}
}
Kostenlos starten →