Introducing Artificial Intelligence in Interpretation of Foetal Cardiotocography: Medical Dataset Curation and...

Area of Science:

• Medical Informatics
• Obstetrics and Gynecology
• Artificial Intelligence in Medicine

Background:

• Cardiotocography (CTG) is crucial for monitoring fetal well-being during labor but suffers from interpretation variability.
• Artificial intelligence (AI) offers potential to enhance CTG analysis and improve clinical decision-making.

Purpose of the Study:

• To develop and validate an AI-assisted interpretation system for Cardiotocography (CTG) to improve the assessment of fetal well-being.
• To create a comprehensive CTG database for training and evaluating AI algorithms for predicting fetal hypoxia.

Main Methods:

• Extracted raw CTG signals and matched them with fetal outcomes (umbilical cord arterial pH, 5-min APGAR score) from 2006-2019.
• Curated a dataset of 6141 paired CTG recordings and clinical data, defining physiological ranges for fetal pH (≥7.15) and APGAR score (≥7).
• Utilized half of the curated dataset to train AI algorithms for predicting fetal hypoxia.

Main Results:

• Successfully created a large, curated database of 6141 CTG recordings with corresponding fetal outcomes, representing the second-largest worldwide.
• Developed AI algorithms trained on this dataset to predict fetal hypoxia based on CTG patterns and clinical data.
• The database serves as a valuable resource for further research in AI-assisted obstetric monitoring.

Conclusions:

• AI-assisted interpretation of CTG has the potential to significantly reduce inter- and intraobserver variability.
• The established CTG database is a key resource for advancing AI applications in predicting fetal hypoxia and improving intrapartal care.
• Further enrichment of the database is planned to enhance the robustness and generalizability of AI models.