Researchers at the Children’s Hospital of Philadelphia (CHOP) have found that the AI chatbot, ChatGPT-4, could be a useful tool in enhancing ultrasound image analysis for differentiating liver pathologies. These findings were published in the journal JMIR AI.
The study explored the integration of AI into medical imaging, particularly using large language models like ChatGPT-4, to improve diagnostic workflows through automated and rapid analysis. The researchers aimed to assess ChatGPT-4’s performance in liver ultrasound radiomics by evaluating its ability to differentiate liver conditions such as fibrosis, steatosis, and normal liver tissue compared to traditional analysis software.
Laith Sultan, MD, MPH, a research scientist and principal investigator with the Center for Pediatric Contrast Ultrasound at CHOP, stated: “With further refinement, ChatGPT-4 can be an effective and scalable tool to enhance patient outcomes in medical imaging.”
In their research, they analyzed 70 grayscale ultrasound images from a preclinical liver disease model including groups with fibrosis, fatty liver, and normal liver. The dataset was processed through both ChatGPT-4 and conventional radiomics methods to evaluate each method’s accuracy, efficiency, and practicality in clinical settings.
The results indicated that ChatGPT-4 performed comparably to traditional radiomics techniques in analyzing liver ultrasound images. It achieved 76% accuracy and 83% sensitivity in distinguishing liver pathologies. Although traditional software had a slightly higher sensitivity of 89%, ChatGPT-4 offered efficiency by processing multiple images concurrently.
Researchers highlighted limitations of ChatGPT-4; its interpretations are based on the data it was trained on and may not cover all possible variations in liver pathology. Therefore, they see ChatGPT-4 as assisting rather than replacing image analysis.
The project received support from the CHOP Department of Radiology under the supervision of Hansel J. Otero, MD; Sudha Anupindi, MD; and Kassa Darge, MD, PhD.
Sultan et al.’s work titled “ChatGPT-4-Driven Liver Ultrasound Radiomics Analysis: Advantages and Drawbacks Compared to Traditional Techniques” was published online on May 19, 2025. DOI:10.2196/68144.
