Hybridized artificial intelligence system for reducing neonatal mortality in Nigeria

Area of Science:

• Artificial Intelligence in Healthcare
• Machine Learning for Medical Diagnosis
• Neonatal Health Informatics

Background:

• Neonatal diseases are a leading cause of death in Nigeria, with high global mortality rates.
• Challenges in early and accurate diagnosis contribute to delayed interventions and increased mortality.
• Southwest Nigeria faces a critical need for improved neonatal disease detection systems.

Purpose of the Study:

• To develop an artificial intelligence (AI) system for detecting multiple neonatal diseases.
• To utilize local Nigerian datasets and advanced machine learning techniques.
• To facilitate early intervention and reduce neonatal mortality in Southwest Nigeria.

Main Methods:

• Collected clinical records from 4,027 neonatal patients across five tertiary hospitals.
• Preprocessed and balanced the dataset using Synthetic Minority Over-sampling Technique (SMOTE).
• Trained and evaluated Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), and a hybrid LSTM-ANN model using accuracy, precision, recall, and F1-score.

Main Results:

• The hybrid LSTM-ANN model achieved 82% accuracy, outperforming ANN (80%) and LSTM (77%).
• Achieved high precision for sepsis (0.90) and birth asphyxia (0.88).
• Statistical tests confirmed the hybrid model's significant superiority over standalone models.

Conclusions:

• The hybrid LSTM-ANN model shows promise as a diagnostic tool for early neonatal disease detection.
• Further external validation and prospective clinical trials are required before widespread clinical adoption.
• This AI approach could significantly impact neonatal mortality rates in resource-limited settings.