A multicentre comparative study of 17 experts and an intelligent computer system for managing labour using the
R D Keith1, S Beckley, J M Garibaldi
1Department of Obstetrics, Postgraduate Medical School, Plymouth, UK.
Summary
An intelligent computer system matched expert performance in interpreting cardiotocograms (CTGs) for labor management, showing higher consistency. This technology could improve CTG interpretation and reduce interventions, highlighting the CTG's potential in clinical practice.
Area of Science:
- Perinatal Medicine
- Artificial Intelligence in Healthcare
- Clinical Decision Support Systems
Background:
- Cardiotocograms (CTGs) are crucial for fetal well-being monitoring during labor.
- Expert interpretation of CTGs can vary, impacting labor management consistency.
- Intelligent computer systems offer potential for standardized analysis of complex medical data.
Purpose of the Study:
- To evaluate if an intelligent computer system can achieve expert-level performance in labor management using CTGs, patient data, and fetal blood sampling.
- To assess the consistency and agreement among experienced clinicians in managing labor based on CTG interpretation.
Main Methods:
- Fifty intrapartum CTG cases with clinical data were independently reviewed twice by an intelligent computer system and 17 experienced fetal monitoring clinicians.
- CTG scoring was performed in 15-minute segments, with cervical dilatation and fetal scalp blood pH estimated upon request.
- Key outcome measures included consistency in scores, agreement on cesarean section recommendations, fetal blood sampling rates, and intervention strategies for various outcomes.
Main Results:
- The intelligent system demonstrated performance comparable to experts, significantly better than chance (67.33%, kappa = 0.31).
- The system exhibited high consistency (99.16%, kappa = 0.98) across independent uses.
- It recommended appropriate interventions, avoiding unnecessary procedures in normal deliveries and identifying birth asphyxia cases effectively, often matching or exceeding expert recommendations.
Conclusions:
- The intelligent computer system's performance was indistinguishable from expert clinicians but demonstrated superior consistency.
- This highlights the potential of intelligent systems to enhance CTG interpretation and reduce interventions.
- The study underscores the effectiveness of CTGs and prompts further investigation into discrepancies between CTG potential and current clinical practice outcomes.


