Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jun 18, 2026

A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation
09:34

A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation

Published on: September 14, 2017

Assessing the Reliability of Large Language Models in Detecting Acute Knee Fractures on Radiographs:A Comparative

Osman Konukoglu1, Murat Kaya1, Baris Can Arslan1

  • 1Department of Radiology, Gaziantep City Hospital, Gaziantep, Türkiye.

Journal of Clinical Practice and Research
|June 17, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Diagnostic Adequacy of Pre-TAVI CTA Compared With ICA and Its Prognostic Association With Coronary Artery Disease.

Journal of medical imaging and radiation oncology·2026
Same author

Association Between Mild Autonomous Cortisol Secretion and Hepatic Fat Fraction in Patients With Adrenal Adenomas: A Chemical Shift MRI Study.

Journal of medical imaging and radiation oncology·2026
Same author

Diagnostic Performance of Diffusion-Weighted Imaging for Evaluation of Breast Lesions: Correlation With Prognostic Factors and Histological Phenotypes.

Journal of clinical ultrasound : JCU·2026
Same author

Ultrasonographic measurements of the uterus and ovaries for the diagnosis of central precocious puberty: A retrospective diagnostic study.

Clinical imaging·2026
Same author

Differentiation of normal pressure hydrocephalus from alzheimer's and Parkinson's Diseases using the splenial angle measured on brain CT: diagnostic performance and reliability study.

Neuroradiology·2026
Same author

Evaluation of Liver Iron Accumulation With Liver Elastography and Apparent Diffusion Coefficient in Children With Thalassaemia Major.

Journal of medical imaging and radiation oncology·2026
Same journal

Predictors of Surgical Success in Parathyroidectomy: A Retrospective Study of Patients Receiving Renal Replacement Therapy with Secondary Hyperparathyroidism.

Journal of clinical practice and research·2026
Same journal

Neuroprotective and Nitric Oxide-Modulating Effects of D-Limonene in a Penicillin-Induced Epilepsy Model in Rats.

Journal of clinical practice and research·2026
Same journal

Do Sexually Transmitted Infections Coexist? Evidence from Human Immunodeficiency Virus and Syphilis Coinfection in a Tertiary Care Center.

Journal of clinical practice and research·2026
Same journal

Limbic System Glioblastoma Extending to the Papez Circuit: A Case Report.

Journal of clinical practice and research·2026
Same journal

The Effect of Myeloid-Derived Suppressor Cells in Graft-versus-Host Disease.

Journal of clinical practice and research·2026
Same journal

Coronary Artery Origin Anomalies on Coronary CT Angiography: A Single-Center Tertiary-Care Cohort.

Journal of clinical practice and research·2026
See all related articles

Large language models (LLMs) showed low accuracy in detecting knee fractures on radiographs, missing many cases. Clinicians, especially radiologists, significantly outperformed LLMs in this diagnostic task.

Area of Science:

  • Radiology
  • Artificial Intelligence in Medicine
  • Medical Imaging Analysis

Background:

  • Acute knee fractures are common injuries requiring accurate and timely diagnosis.
  • Radiographs are a primary imaging modality for evaluating knee injuries.
  • The diagnostic performance of advanced AI tools, specifically large language models (LLMs), in this context is under investigation.

Purpose of the Study:

  • To evaluate the diagnostic accuracy and reliability of closed-source, multimodal LLMs (ChatGPT-4o, ChatGPT-4.5, Gemini 2.5 Pro) in detecting acute knee fractures on radiographs.
  • To compare the performance of these LLMs against an emergency medicine specialist and a radiologist.

Main Methods:

  • A retrospective study of 252 patients with knee radiography and CT scans.
  • Fracture status confirmed by CT and radiologist review.
Keywords:
FractureX-rayknee traumalarge language modelsradiology

Related Experiment Videos

Last Updated: Jun 18, 2026

A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation
09:34

A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation

Published on: September 14, 2017

  • Independent assessment of radiographs by an emergency medicine specialist, a radiologist, and three LLMs.
  • Evaluation of diagnostic performance using sensitivity, specificity, accuracy, and AUC; reliability assessed with Cohen's kappa and McNemar's tests.

    • Main Results:

    • LLMs exhibited low sensitivity (10.3%-37.9%) and moderate overall accuracy (72%-77%).
    • The radiologist achieved 92.1% accuracy with high sensitivity (77.6%) and specificity (96.4%).
    • The emergency medicine specialist achieved 83.7% accuracy; clinicians significantly outperformed LLMs (p<0.05).

    Conclusions:

    • Closed-source LLMs demonstrated inferior performance compared to clinicians in diagnosing acute knee fractures on radiographs, with a high risk of missed fractures.
    • LLMs may assist in triaging by identifying normal cases but are insufficient for standalone diagnostic use.
    • Further research into AI-assisted fracture detection is warranted, focusing on improving sensitivity and reliability.