Artificial intelligence-large language models (AI-LLMs) for reliable and accurate cardiotocography (CTG) interpretation in obstetric practice

Khanisyah Erza Gumilar ^1,2, Manggala Pasca Wardhana ³, Muhammad Ilham Aldika Akbar ²

¹Graduate Institute of Biomedical Science, China Medical University, Taichung, Taiwan.
²Department of Obstetrics and Gynecology, Universitas Airlangga Hospital - Faculty of Medicine, Universitas Airlangga, Surabaya, Indonesia.
³Department of Obstetrics and Gynecology, Dr. Soetomo General Hospital - Faculty of Medicine, Universitas Airlangga, Surabaya, Indonesia.
⁴Department of Obstetrics and Gynecology, Dr. Ramelan Naval Hospital, Surabaya, Indonesia.
⁵Department of Obstetrics and Gynecology, Dr. Mohamad Soewandhie Hospital, Surabaya, Indonesia.
⁶Department of Obstetrics and Gynecology, Department of Obstetrics and Gynecology, Udayana University Hospital, Denpasar, Indonesia.
⁷Department of Obstetrics and Gynecology, Bandung Kiwari General Hospital, Bandung, Indonesia.
⁸Department of Public Health, China Medical University, Taichung, Taiwan.
⁹School of Chinese Medicine, China Medical University, Taichung, Taiwan.
¹⁰Department of Obstetrics and Gynecology, Taichung Veteran General Hospital, Taichung, Taiwan.
¹¹Institute of Biochemistry and Molecular Biology and Research Center for Cancer Biology, China Medical University, Taichung, Taiwan.

⁰Graduate Institute of Biomedical Science, China Medical University, Taichung, Taiwan.

Computational and Structural Biotechnology Journal +

|

April 10, 2025

View abstract on PubMed

Summary

ChatGPT-4o demonstrated superior performance in cardiotocography interpretation, outperforming other AI-LLMs and junior doctors. This advanced AI tool closely matched senior doctor performance, showing promise for improving obstetric care.

Area Of Science

Medical Imaging
Artificial Intelligence
Obstetrics

Background

Accurate cardiotocography (CTG) interpretation is crucial for monitoring fetal well-being during pregnancy and labor.
Advanced artificial intelligence (AI) tools, specifically AI-large language models (AI-LLMs), have the potential to improve CTG interpretation accuracy.
The clinical utility and performance of AI-LLMs in CTG interpretation require thorough evaluation.

Purpose Of The Study

To assess the performance of three AI-LLMs (ChatGPT-4o, Gemini Advanced, Copilot) in interpreting CTG images.
To compare the AI-LLMs' interpretations with those of junior (JHDs) and senior human doctors (SHDs).
To evaluate the reliability of AI-LLMs in clinical decision-making for obstetric care.

Main Methods

Seven CTG images were interpreted by three AI-LLMs, five SHDs, and five JHDs.
Expert evaluations were conducted by five blinded maternal-fetal medicine specialists using a Likert scale for relevance, clarity, depth, focus, and coherence.
Statistical analysis was performed to compare the homogeneity of expert ratings and group performances.

Main Results

ChatGPT-4o achieved a score of 77.86, outperforming Gemini Advanced (57.14), Copilot (47.29), and JHDs (61.57).
ChatGPT-4o's performance (77.86) closely approached that of SHDs (80.43), with no statistically significant difference (p > 0.05).
ChatGPT-4o particularly excelled in the depth parameter and was only marginally inferior to SHDs in other assessed parameters.

Conclusions

ChatGPT-4o demonstrated superior performance among evaluated AI-LLMs in CTG interpretation.
ChatGPT-4o surpassed the performance of junior doctors and closely matched senior doctors.
AI-LLMs, exemplified by ChatGPT-4o, show significant potential as assistive tools for obstetricians, aiming to enhance diagnostic accuracy and improve obstetric patient care.

Keywords:

Artificial intelligence-large language models (AI-LLMs) Cardiotocography (CTG) ChatGPT Copilot Fetal monitoring Gemini Obstetrics