Artificial Intelligence for Early Pregnancy Risk Prediction in Nigeria: A Multi-Modal Machine Learning Approach

Early detection of pregnancy-related risks remains a major maternal health challenge in Nigeria and other low- and middle-income countries, where preventable maternal and fetal morbidity and mortality are worsened by limited specialist care, late antenatal booking, and fragmented health data systems. This study develops and evaluates a multimodal artificial intelligence framework for early pregnancy risk prediction in Nigeria using heterogeneous maternal health data, including clinical records, socio-demographic characteristics, obstetric and medical history, laboratory findings, and lifestyle-related variables. Three machine learning models—Random Forest, Gradient Boosting, and Deep Neural Networks—were trained and compared for first-trimester pregnancy risk classification. To enhance clinical interpretability and support transparent decision-making, Shapley Additive Explanations (SHAP) were applied to identify the most influential predictors contributing to risk stratification. The findings indicate that integrating multiple data modalities improves predictive performance compared with models based on isolated data sources, demonstrating the value of comprehensive maternal health profiling for early risk assessment. The results further suggest that explainable AI can support clinicians by providing interpretable, data-driven insights for timely intervention, antenatal care planning, and prioritization of high-risk pregnancies. This study contributes a context-sensitive, scalable, and ethically informed AI framework for improving maternal health service delivery in resource-constrained Nigerian healthcare settings and across Sub-Saharan Africa.