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 25, 2026

Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis
06:03

Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis

Published on: February 6, 2020

Application of LLMs in CAD-RADS Classification and Patient Management.

Piotr Tarkowski1,2,3, Giuseppe Muscogiuri3,4,5, Davide Casartelli3

  • 1Department of Radiology and Nuclear Medicine, University Hospital No 4 of Lublin, Lublin, Poland.

Echocardiography (Mount Kisco, N.Y.)
|June 24, 2026
PubMed
Summary

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

Efficacy and safety of transcatheter left atrial appendage closure guided by transoesophageal echocardiography using standard and micro TEE probes: a single-center retrospective study.

Frontiers in cardiovascular medicine·2026
Same author

Levator scapulae muscle variability: pathology or anatomical variant? A narrative review with pragmatic classification and imaging relevance.

Folia morphologica·2026
Same author

Deep Learning-Based Metal Artifact Reduction in Cardiac Computed Tomography: A Preliminary Study Enabling Radiomic Analysis in Patients with Implantable Defibrillators.

Journal of imaging informatics in medicine·2026
Same author

Prognostic value of the baseline CAVIAR response score after cardiac resynchronization therapy: 10-year outcomes from the Markers and Response to CRT study.

Heart rhythm·2026
Same author

Reprogramming the Immunosuppressive Microenvironment in Glioblastoma Through Oncolytic Herpes Simplex Virus Therapy: A Systematic Review.

Cells·2026
Same author

Association of Relative Pericoronary Adipose Tissue Attenuation with Coronary Artery Calcification Severity.

Medicina (Kaunas, Lithuania)·2026
This summary is machine-generated.

Large language models (LLMs) show promise in assigning Coronary Artery Disease-Reporting and Data System (CAD-RADS) scores, but struggle with higher grades and patient management recommendations. Further development is needed for clinical reliability.

Area of Science:

  • Artificial Intelligence in Medical Imaging
  • Natural Language Processing in Healthcare
  • Cardiovascular Diagnostics

Background:

  • Coronary CT angiography (CCTA) reports are crucial for diagnosing Coronary Artery Disease (CAD).
  • Standardized scoring systems like CAD-RADS improve reporting consistency.
  • Large language models (LLMs) are increasingly explored for medical applications.

Purpose of the Study:

  • To assess the accuracy of four LLMs in assigning CAD-RADS scores from CCTA reports.
  • To evaluate LLM-generated patient management recommendations based on CCTA findings.
  • To compare LLM performance across different models and prompt perspectives (cardiologist vs. radiologist).

Main Methods:

  • Four LLMs (ChatGPT 4o, Claude 3.7, DeepSeek, Gemini 2.5 Pro) analyzed synthetic CCTA reports.
Keywords:
CADRADSartificial intelligencecardiac computed tomographycoronary artery diseasemyocardial ischemia

More Related Videos

Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
07:13

Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model

Published on: April 18, 2025

Related Experiment Videos

Last Updated: Jun 25, 2026

Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis
06:03

Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis

Published on: February 6, 2020

Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
07:13

Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model

Published on: April 18, 2025

  • Prompts simulated cardiologist and radiologist viewpoints.
  • Agreement with human standards was measured using Cohen's kappa, Fleiss' kappa, and Krippendorff's alpha for scoring and management recommendations.

    • Main Results:

    • Claude 3.7 and Gemini 2.5 Pro demonstrated near-perfect agreement for CAD-RADS scoring.
    • ChatGPT showed high agreement as a radiologist but lower as a cardiologist.
    • All LLMs struggled with higher-grade stenosis (CAD-RADS 4A/4B) and management recommendations, with significant variability observed.
    • Prompt identity influenced LLM performance.

    Conclusions:

    • LLMs reliably score lower-grade CAD-RADS categories (0-2).
    • Performance significantly decreases for higher-grade stenosis (4A/4B) and non-diagnostic reports, posing potential patient risks.
    • Current LLM capabilities for generating dependable clinical management recommendations are limited.