Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

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

Further Methodological and Clinical Reflections-Reply.

JAMA otolaryngology-- head & neck surgery·2026

Same author

Champions and early adopters in an implementation-effectiveness study of shared decision making: implications for interpersonal strategies.

Implementation science communications·2026

Same author

Encounter tool for Shared Decision Making about Adjuvant Treatment of Lung Cancer: Randomized Clinical Trial.

Research square·2026

Same author

Comparative effectiveness of sulfonylureas on kidney outcomes in adults with type 2 diabetes and moderate cardiovascular risk: a target trial emulation.

BMJ open diabetes research & care·2026

Same author

Comparative Effectiveness of Individual Sodium Glucose Transporter 2 Inhibitors on Cardiovascular Outcomes in Type 2 Diabetes With Moderate Cardiovascular Risk: Emulation of a Target Trial.

Journal of the American Heart Association·2026

Same author

Clinician in the loop: a flawed solution for AI oversight.

BMJ (Clinical research ed.)·2026

Search research articles

Related Experiment Video

Updated: Jun 16, 2025

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

ChatGPT-4's Accuracy in Estimating Thyroid Nodule Features and Cancer Risk From Ultrasound Images.

Esteban Cabezas¹, David Toro-Tobon², Thomas Johnson³

¹Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic, Rochester, Minnesota.

Endocrine Practice : Official Journal of the American College of Endocrinology and the American Association of Clinical Endocrinologists

|March 26, 2025

Summary

AI models GPT-4 and GPT-4o show potential for thyroid nodule ultrasound analysis but require further refinement. Their current performance is suboptimal, especially for higher-risk nodules, necessitating more validation before clinical use.

Keywords:

TI-RADS artificial intelligence chatbot large language models thyroid nodule ultrasound

More Related Videos

Author Spotlight: Integrating Ultrasound Imaging with Biochemical Markers for Thyroid Disease Diagnosis

Author Spotlight: Integrating Ultrasound Imaging with Biochemical Markers for Thyroid Disease Diagnosis

Published on: February 9, 2024

Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules

Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules

Published on: May 19, 2023

Related Experiment Videos

Last Updated: Jun 16, 2025

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

Author Spotlight: Integrating Ultrasound Imaging with Biochemical Markers for Thyroid Disease Diagnosis

Author Spotlight: Integrating Ultrasound Imaging with Biochemical Markers for Thyroid Disease Diagnosis

Published on: February 9, 2024

Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules

Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules

Published on: May 19, 2023

Area of Science:

Artificial Intelligence in Medical Imaging
Radiology and Diagnostic Imaging
Endocrinology and Thyroidology

Background:

Thyroid nodules are common, requiring accurate characterization for appropriate management.
The American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) provides a standardized framework for classifying thyroid nodules.
AI tools are being explored to assist in medical image analysis and improve diagnostic efficiency.

Purpose of the Study:

To evaluate the diagnostic performance of GPT-4 and GPT-4o in identifying thyroid nodule features and TI-RADS categories from ultrasound images.
To compare the accuracy of these AI models against expert radiologist assessments.

Main Methods:

A comparative validation study using 202 thyroid ultrasound images from open-access databases.
Expert radiologists established a reference standard for TI-RADS features and categories for both complete and cropped images.
GPT-4 and GPT-4o were prompted to analyze images, and their outputs were compared to the established reference standard.

Main Results:

GPT-4 exhibited high specificity but low sensitivity for most TI-RADS categories on complete images, leading to variable overall accuracy.
For low-risk nodules, GPT-4 achieved 25.0% sensitivity and 99.5% specificity. In moderately suspicious categories, sensitivity was 75% with 22.2% specificity.
The AI models struggled with identifying specific features like isoechoic echogenicity and echogenic foci, and performance slightly decreased with cropped images.

Conclusions:

GPT-4 and GPT-4o demonstrate potential for enhancing thyroid nodule triage efficiency.
Current AI model performance is suboptimal, particularly for higher-risk thyroid nodules.
Further research, refinement, and validation are crucial for the clinical implementation of these AI tools in thyroid imaging.