Machine Learning Approaches for Predicting Neurodevelopmental Disorder Risk Among Neonates with Perinatal Asphyxia: A Review

Perinatal asphyxia remains a leading contributor to neonatal mortality and morbidity in Nigeria, with survivors facing a substantially elevated risk of long-term neurodevelopmental impairment, including cerebral palsy, intellectual disability, and epilepsy. Early identification of at-risk neonates is essential for timely intervention, yet conventional clinical assessment tools are often insufficient for early, individualized risk stratification, particularly in low-resource settings. This review synthesizes current evidence on the burden and pathophysiology of perinatal asphyxia, its established link to neurodevelopmental disorders, and the growing application of machine learning (ML) techniques, notably logistic regression and random forest, in neonatal risk prediction. A structured review of empirical and grey literature indicates that, although ML has shown promise for disease prediction and risk stratification in broader paediatric and obstetric contexts, no existing study has developed or validated an ML-based model specifically for predicting neurodevelopmental disorder risk among Nigerian neonates with perinatal asphyxia. Most available Nigerian evidence is hospital-based, retrospective, and lacks a predictive or nationally representative dimension. This review proposes a conceptual machine learning pipeline, comprising data acquisition, preprocessing, feature engineering, model training, and evaluation, as a framework for addressing this gap. The findings underscore the potential of interpretable and ensemble-based ML models to support early clinical decision-making and improve neurodevelopmental outcomes in Nigeria and comparable low- and middle-income settings.