A Novel Ensemble-Based AI Framework for Early Prediction of Monogenic Type 1 Diabetes in Neonates Using Maternal and Pregnancy Health Indicators

While monogenic Type 1 diabetes in neonates is uncommon, it can critically endanger a child’s health if not diagnosed promptly. This study aims to develop an ensemble-based predictive framework utilizing AI for identifying neonatal monogenic diabetes risks using maternal and pregnancy-related health indicators. By employing a publicly available dataset, we reconstructed neonatal outcomes using AI-based pattern recognition. The proposed model utilizes Decision Trees (DT), Random Forests (RF), Gradient Boosted Trees (GBT), and K-Nearest Neighbors (KNN) in a soft-voting ensemble framework. As our results indicate, ensemble methods outperformed individual classifiers, with ensemble approaches yielding higher accuracy as well as improved generalization. Overall, the framework can support clinicians in off-screening at-risk neonates, guiding proactive clinical action and tailored care after birth.