Predicting adverse perinatal outcomes among gestational diabetes complicated pregnancies using neural network...

Area of Science:

• Perinatal medicine
• Artificial intelligence in healthcare
• Maternal-fetal medicine

Background:

• Gestational diabetes mellitus (GDM) is a significant complication affecting maternal and neonatal health.
• Predicting adverse neonatal outcomes in GDM pregnancies remains a clinical challenge.
• Early identification of high-risk pregnancies is crucial for timely intervention.

Purpose of the Study:

• To develop and validate a neural network model for predicting adverse neonatal outcomes in pregnancies with GDM.
• To identify key predictive factors for adverse neonatal outcomes in GDM pregnancies.
• To improve risk stratification and management strategies for GDM.

Main Methods:

• Utilized an XGBoost-based neural network model implemented in Python 3.6 with Keras and TensorFlow.
• Sourced data from 452 GDM-diagnosed individuals delivering between 2012-2016.
• Included pregnancy parameters, maternal demographics, glucose tolerance test results, and glycemic control data.

Main Results:

• The model achieved 82% prediction accuracy at GDM diagnosis and 91% at delivery.
• Adverse neonatal outcomes occurred in 29% of the cohort.
• Key predictors included maternal age, pre-pregnancy BMI, and 3-hour oral glucose tolerance test results.

Conclusions:

• The developed neural network model demonstrates significant potential for predicting adverse neonatal outcomes in GDM pregnancies.
• This predictive tool can aid clinicians in identifying high-risk pregnancies for improved neonatal care.
• Further validation and implementation could enhance management protocols for GDM.