Prediction of preterm birth using artificial intelligence: a systematic review

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Neonatal Health

Background:

• Preterm birth is the primary cause of neonatal mortality and morbidity.
• Accurate prediction of preterm birth remains a significant clinical challenge.

Purpose of the Study:

• To systematically review and analyze the current state of artificial intelligence (AI) research for preterm birth prediction.
• To clarify the predictive values and accuracy of various AI models and data types used in preterm birth prediction.

Main Methods:

• A systematic review of three major databases (PubMed, Web of Science, Scopus) was conducted in August 2020.
• Keywords included 'artificial intelligence,' 'deep learning,' 'machine learning,' and 'neural network' combined with 'preterm birth'.
• Twenty-two relevant publications from 2010-2020 were included in the analysis.

Main Results:

• Electrohysterogram images were most frequently used for prediction, followed by biological profiles, metabolic panels, and cervical ultrasound images.
• Studies utilizing metabolic panels and electrohysterogram images demonstrated better prediction accuracy.
• Most studies used small datasets (around 100 cases), with only three using large databases (>100,000 cases).

Conclusions:

• AI, including deep learning and machine learning, can potentially achieve accurate preterm birth prediction using clinical data.
• The accuracy of AI models is influenced by the type of data used, with metabolic panels and electrohysterograms showing promise.
• Future research requires significantly larger datasets to train robust AI models for reliable preterm birth prediction.