Artificial Intelligence-Large Language Models (AI-LLMs) for Reliable and Accurate Cardiotocography (CTG) Interpretation in Obstetric Practice

Abstract

Abstract BACKGROUND Accurate interpretation of Cardiotocography (CTG) is a critical tool for monitoring fetal well-being during pregnancy and labor, providing crucial insights into fetal heart rate and uterine contractions. Advanced artificial intelligence (AI) tools such as AI-Large Language Models (AI-LLMs) may enhance the accuracy of CTG interpretation, leading to better clinical outcomes. However, this potential has not yet been examined and reported yet. OBJECTIVE This study aimed to evaluate the performance of three AI-LLMs (ChatGPT-4o, Gemini Advance, and Copilot) in interpreting CTG images, comparing their performance to junior and senior human doctors, and assessing their reliability in assisting clinical decisions. STUDY DESIGN: Seven CTG images were evaluated by three AI-LLMs, five senior doctors (SHD), and five junior doctors (JHD) and rated by five maternal-fetal medicine (MFM) experts (raters) using five parameters (relevance, clarity, depth, focus, and coherence). The raters were blinded to the source of interpretations, and a Likert scale was used to score the performance of each system. Statistical analysis assessed the homogeneity of expert ratings and the comparative performance of AI-LLMs and doctors. RESULTS ChatGPT-4o outperformed the other AI models with a score of 77.86, much higher than Gemini Advance (57.14) and Copilot (47.29), as well as the junior doctors (JHD; 61.57). CG4o’s performance (77.86) was only slightly below that of the senior doctor (SHD; 80.43), with no statistically significant differences between CG4o and SHD (p>0.05). Meanwhile, CG4o had the greatest score in the “depth” category, while the other four parameters were only marginally behind SHD. CONCLUSION CG4o demonstrated outstanding performance in CTG interpretation, surpassing junior doctors and other AI-LLMs, while senior doctors remain superior in all groups. AI-LLMs, particularly CG4o, showed promising potential as valuable tools in clinical practice to assist obstetricians, enhance diagnostic accuracy, and improve patient care.

Ähnliche Arbeiten