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

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiomics

Background:

• Artificial intelligence (AI) is revolutionizing medical imaging.
• Large language models (LLMs) like ChatGPT-4 are emerging as tools for automated image interpretation.
• The application of ChatGPT-4 in liver ultrasound radiomics is largely unexplored.

Purpose of the Study:

• To evaluate ChatGPT-4's capability in liver ultrasound radiomics.
• To assess its ability to differentiate fibrosis, steatosis, and normal liver tissue.
• To compare its performance against conventional image analysis software.

Main Methods:

• Seventy grayscale ultrasound images from a preclinical liver disease model were analyzed.
• ChatGPT-4 extracted texture features, compared with Interactive Data Language (IDL).
• Statistical analyses (ANOVA, logistic regression) assessed feature significance and diagnostic performance.

Main Results:

• ChatGPT-4 identified nine key textural features with significant differences across liver conditions (p < 0.05).
• It achieved 76% accuracy and 83% sensitivity in classifying liver disease, with high AUC values for specific conditions (e.g., 0.97 for steatosis).
• ChatGPT-4 demonstrated 40% faster processing than IDL, indicating high-throughput potential.

Conclusions:

• ChatGPT-4 shows high feasibility for ultrasound radiomics, despite slightly lower sensitivity than IDL.
• It offers faster processing, high-throughput analysis, and automated multi-image evaluation.
• Further refinement is needed to enhance feature reproducibility and diagnostic accuracy for AI-driven imaging workflows.