Related Experiment Video
Updated: Dec 10, 2025
06:56
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
2.7K
Concordance analysis of intrapartum cardiotocography between physicians and artificial intelligence-based technique
Li-Chun Liu1,2, Ya-Hui Tsai3,4, Yu-Ching Chou5
1Department of Obstetrics and Gynecology, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan, ROC.
Journal of the Chinese Medical Association : JCMA
|August 29, 2020
Summary
A novel artificial intelligence method using fully convolutional networks (FCNs) shows promise for automated electronic fetal monitoring (EFM) analysis, accurately recognizing fetal heart rate patterns and aiding in the assessment of nonreassuring fetal status.
Area of Science:
- Medical Imaging and Artificial Intelligence
- Obstetrics and Gynecology
- Fetal Monitoring Technology
Background:
- Cardiotocography is a standard electronic fetal monitoring (EFM) technique for assessing fetal well-being.
- Data-driven approaches, including artificial intelligence (AI), offer potential for automating EFM analysis.
- This study explores a novel AI method for EFM evaluation and recognition of nonreassuring fetal status.
Purpose of the Study:
- To evaluate a novel artificial intelligence method based on fully convolutional networks (FCNs) for electronic fetal monitoring (EFM) trace recognition.
- To assess the potential role of FCNs in the evaluation of nonreassuring fetal status.
- To compare the performance of the FCN model against clinical practice in EFM analysis.
Main Methods:
- Retrospective collection of 3239 EFM labor records from 292 deliveries.
- Analysis of EFM data using a fully convolutional network (FCN) model.
- Comparison of FCN model results with clinical practice assessments and neonatal Apgar scores.
Main Results:
- The FCN model demonstrated physician-level recognition of EFM traces, achieving a Cohen's kappa of 0.525 and an AUC of 0.892 for six fetal heart rate (FHR) categories.
- The FCN model exhibited significantly higher sensitivity (0.528 vs 0.132) in predicting fetal compromise compared to clinical practice.
- The FCN model had a higher false-positive rate (0.632 vs 0.012) than clinical practice.
Conclusions:
- Fully convolutional networks (FCNs) represent a modern technique with potential utility in electronic fetal monitoring (EFM) trace recognition.
- The developed FCN model shows competitive performance in identifying fetal heart rate (FHR) patterns.
- The AI model holds potential for assisting in the evaluation of nonreassuring fetal status, warranting further investigation.
