EUVIMEDEUVIMED entdecken
Vergleichsstudie

Large Language Models for Diagnosing Focal Liver Lesions From CT / MRI Reports: A Comparative Study With Radiologists

Liuji Sheng, Yidi Chen iD, Hong Wei, Feng Che, Yingyi Wu iD, Qin Qin, Chongtu Yang, Yanshu Wang, Jingwen Peng, Mustafa R Bashir, Maxime Ronot iD, Bin Song iD und 1 weitere

Liver International · 2025

DOI: 10.1111/liv.70115 ↗

Kurz gefasst

Vergleichsstudie in der Zeitschrift Liver International, 2025. Verfasst von Liuji Sheng und 12 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiology and Functional and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Si…. Hauptthemen laut MEDLINE: Magnetic Resonance Imaging, Tomography, X-Ray Computed, Liver Neoplasms. Die Arbeit wurde bisher 17-mal zitiert (Crossref) und nennt 27 Literaturangaben. Gefördert von: National Health Commission Capacity Building and Continuing Education Center, National Natural Science Foundation of China, The Science and Technology Department of Hainan Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background & Aims Whether large language models (LLMs) could be integrated into the diagnostic workflow of focal liver lesions (FLLs) remains unclear. We aimed to investigate two generic LLMs (ChatGPT‐4o and Gemini) regarding their diagnostic accuracies referring to the CT/MRI reports, compared to and combined with radiologists of different experience levels.

Methods

From April 2022 to April 2024, this single‐center retrospective study included consecutive adult patients who underwent contrast‐enhanced CT/MRI for single FLL and subsequent histopathologic examination. The LLMs were prompted by clinical information and the “findings” section of radiology reports three times to provide differential diagnoses in the descending order of likelihood, with the first considered the final diagnosis. In the research setting, six radiologists (three junior and three middle‐level) independently reviewed the CT/MRI images and clinical information in two rounds (first alone, then with LLM assistance). In the clinical setting, diagnoses were retrieved from the “impressions” section of radiology reports. Diagnostic accuracy was investigated against histopathology.

Results

228 patients (median age, 59 years; 155 males) with 228 FLLs (median size, 3.6 cm) were included. Regarding the final diagnosis, the accuracy of two‐step ChatGPT‐4o (78.9%) was higher than single‐step ChatGPT‐4o (68.0%, p < 0.001) and single‐step Gemini (73.2%, p = 0.004), similar to real‐world radiology reports (80.0%, p = 0.34) and junior radiologists (78.9%–82.0%; p ‐values, 0.21 to > 0.99), but lower than middle‐level radiologists (84.6%–85.5%; p ‐values, 0.001 to 0.02). No incremental diagnostic value of ChatGPT‐4o was observed for any radiologist ( p ‐values, 0.63 to > 0.99).

Conclusion

Two‐step ChatGPT‐4o showed matching accuracies to real‐world radiology reports and junior radiologists for diagnosing FLLs but was less accurate than middle‐level radiologists and demonstrated little incremental diagnostic value.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Magnetic Resonance ImagingTomography, X-Ray ComputedLiver NeoplasmsHumansMaleFemaleRetrospective StudiesMiddle AgedAgedRadiologistsAdultDiagnosis, DifferentialLiverLarge Language Models

Stichwörter der Autor:innen

diagnosisfocal liver lesionslarge language model

Autor:innen und Institutionen

  • Department of Radiology and Functional and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, Sichuan, China.
  • Department of Radiology, The First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi, China.
  • College of Nursing, North Sichuan Medical College, Nanchong, Sichuan, China.
  • Division of Gastroenterology, Department of Radiology, Center for Advanced Magnetic Resonance in Medicine, Department of Medicine, Duke University Medical Center, Durham, North Carolina, USA.
  • Université Paris Cité, UMR 1149, CRI, Paris & Service de Radiologie, Hôpital Beaujon, APHP. Nord, Clichy, France.
  • Department of Radiology, Sanya People's Hospital, Sanya, Hainan, China.

ORCID-Profile: Yidi Chen, Yingyi Wu, Maxime Ronot, Bin Song, Hanyu Jiang

Förderung

  • National Health Commission Capacity Building and Continuing Education Center
  • National Natural Science Foundation of China
  • The Science and Technology Department of Hainan Province
  • the China Postdoctoral Science Foundation
  • Development Project of Hainan Provincial Clinical Medical Center; and Post-doctoral Station Development Project of Sanya
  • Natural Science Foundation of Sichuan Province

Zitieren

APA

Sheng, L., Chen, Y., Wei, H., Che, F., Wu, Y., Qin, Q., Yang, C., Wang, Y., Peng, J., Bashir, M. R., Ronot, M., Song, B., & Jiang, H. (2025). Large Language Models for Diagnosing Focal Liver Lesions From CT / MRI Reports: A Comparative Study With Radiologists. Liver International. https://doi.org/10.1111/liv.70115

Vancouver

Sheng L, Chen Y, Wei H, Che F, Wu Y, Qin Q, et al. Large Language Models for Diagnosing Focal Liver Lesions From CT / MRI Reports: A Comparative Study With Radiologists. Liver International. 2025. doi:10.1111/liv.70115

BibTeX
@article{Sheng2025,
  author = {Sheng, Liuji and Chen, Yidi and Wei, Hong and Che, Feng and Wu, Yingyi and Qin, Qin and Yang, Chongtu and Wang, Yanshu and Peng, Jingwen and Bashir, Mustafa R and Ronot, Maxime and Song, Bin and Jiang, Hanyu},
  title = {Large Language Models for Diagnosing Focal Liver Lesions From CT / MRI Reports: A Comparative Study With Radiologists},
  journal = {Liver International},
  year = {2025},
  doi = {10.1111/liv.70115}
}
Kostenlos starten →