Title: A Proposed Machine Learning-Driven Framework for Multi-Disease Classification in Neonatal Morbidity Prediction Using LSTM Networks in Nigeria
Authors: Charity Segun ODEYEMI, and Olatayo Moses OLANIYAN
Volume: 10
Issue: 2
Pages: 1-6
Publication Date: 2026/02/28
Abstract:
Neonatal morbidity and mortality are significant public health challenges in many low- and middle-income countries, with Nigeria facing problems such as respiratory distress syndrome, pneumonia, jaundice, prematurity, birth asphyxia, congenital anomalies, neonatal infections, and hemolytic diseases, among others, contributing to neonatal mortality. During the initial 28 days of life, the neonate makes significant physiological adaptations from intrauterine to postnatal life, during which the immature organ systems are vulnerable to minor pathological infections. However, while early and effective detection is essential in improving survival outcomes, existing detection methods, which are based on clinical observation, have limitations of subjectivity and delay in diagnosis. Although some studies have employed machine learning in disease detection, they have focused mainly on adults and disease classification, with little emphasis placed on data sets relevant to the Nigerian population. In this study, a framework is proposed that utilizes deep learning and its subset, Long Short-Term Memory (LSTM) networks, which will facilitate multi-disease classification in neonates based on data sets from tertiary healthcare institutions in Southwestern Nigeria, with the aim of more accurately and timely identifying diseases that commonly occur in neonates. The integration of artificial intelligence in neonatal disease detection is intended to improve clinical decisions and lower mortality rates among this vulnerable group of patients in resource-constrained settings.