Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Accepted for/Published in: JMIR AI

Date Submitted: Oct 29, 2024
Date Accepted: May 18, 2025
Date Submitted to PubMed: May 19, 2025

The final, peer-reviewed published version of this preprint can be found here:

ChatGPT-4–Driven Liver Ultrasound Radiomics Analysis: Diagnostic Value and Drawbacks in a Comparative Study

Sultan L, Venkatakrishna SSB, Anupindi S, Andronikou S, Acord M, Otero H, Darge K, Sehgal C, Holmes J

ChatGPT-4–Driven Liver Ultrasound Radiomics Analysis: Diagnostic Value and Drawbacks in a Comparative Study

JMIR AI 2025;4:e68144

DOI: 10.2196/68144

PMID: 40388838

PMCID: 12260471

ChatGPT-4-Driven Liver Ultrasound Radiomics Analysis: Advantages and Drawbacks Compared to Traditional Techniques

  • Laith Sultan; 
  • Shyam Sunder B Venkatakrishna; 
  • Sudha Anupindi; 
  • Savvas Andronikou; 
  • Michael Acord; 
  • Hansel Otero; 
  • Kassa Darge; 
  • Chandra Sehgal; 
  • John Holmes

ABSTRACT

Background:

The integration of artificial intelligence (AI) into medical imaging, particularly with large language models like ChatGPT-4, has the potential to advance diagnostic workflows by enabling automated and rapid analysis. ChatGPT-4's intuitive interface and ability to interpret complex queries position it as a transformative tool for medical image analysis. .

Objective:

This study aims to evaluate ChatGPT-4's performance in liver ultrasound radiomics, assessing its ability to differentiate liver conditions—specifically fibrosis, steatosis, and normal liver tissue—compared to traditional analysis software.

Methods:

Seventy grayscale ultrasound images from a preclinical liver disease model—including groups with fibrosis, fatty liver, and normal liver—were analyzed. Key texture features were extracted using ChatGPT-4 and compared with those obtained from conventional image analysis software (Interactive Data Language, IDL). Statistical significance of texture features distinguishing between liver conditions was assessed using one-way ANOVA. Logistic regression models were fit to evaluate the diagnostic performance of both individual and combined features.

Results:

ChatGPT-4 demonstrated robust diagnostic capabilities, achieving 76% accuracy and 83% sensitivity in distinguishing liver pathologies. Several texture features significantly differentiated between liver conditions. Although the IDL software achieved a slightly higher sensitivity of 89%, ChatGPT-4 offered the added benefit of concurrent processing of multiple images, enhancing analysis efficiency.

Conclusions:

While ChatGPT-4 exhibited slightly lower sensitivity and lacked manual control over region-of-interest selection, it showed promise as a viable tool for ultrasound image analysis, effectively differentiating liver pathologies. Its capability to process multiple images simultaneously could streamline diagnostic workflows and alleviate radiologist workload. With further refinement, ChatGPT-4 has the potential to enhance patient outcomes in medical imaging


 Citation

Please cite as:

Sultan L, Venkatakrishna SSB, Anupindi S, Andronikou S, Acord M, Otero H, Darge K, Sehgal C, Holmes J

ChatGPT-4–Driven Liver Ultrasound Radiomics Analysis: Diagnostic Value and Drawbacks in a Comparative Study

JMIR AI 2025;4:e68144

DOI: 10.2196/68144

PMID: 40388838

PMCID: 12260471

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.