Machine learning on cardiotocography data to classify fetal outcomes: A scoping review

Area of Science:

• Obstetrics and Gynecology
• Medical Imaging and Signal Processing
• Artificial Intelligence in Medicine

Background:

• Intrapartum fetal hypoxia during labor poses risks to newborns, including neurological injury or death.
• Cardiotocography (CTG) monitoring is standard for detecting fetal hypoxia but suffers from poor positive predictive value and interpretation variability.
• Machine learning (ML) offers potential for more accurate fetal hypoxia prediction from CTG data.

Purpose of the Study:

• To review machine learning (ML) techniques applied to cardiotocography (CTG) classification for fetal hypoxia detection.
• To identify research gaps hindering the clinical implementation of ML tools for intrapartum fetal monitoring.

Main Methods:

• A systematic search of PubMed, EMBASE, and IEEE Xplore was conducted using relevant keywords.
• Studies were screened using Preferred Reporting Items for Systematic Review and Meta-Analysis for Scoping Reviews (PRISMA-ScR) guidelines.
• 36 studies employing signal processing and ML for CTG classification were included.

Main Results:

• Most included studies utilized open-access CTG databases and fetal metabolic acidosis as a hypoxia benchmark.
• Varied pH levels for hypoxia classification and limited, non-diverse datasets were identified as significant concerns.
• Diverse ML algorithms were applied to processed CTG signals for classification.

Conclusions:

• Machine learning models show potential for improved prediction of fetal hypoxia from CTG.
• Clinical implementation requires more diverse datasets, standardized hypoxia benchmarks, and enhanced ML algorithms and features.