EARLY DETECTION OF CARDIAC ARREST IN NEWBORNS: A MACHINE LEARNING AND STATISTICAL MODEL APPROACH IN THE CARDIAC INTENSIVE CARE UNIT

Abstract

Cardiac arrest in newborns is a critical medical emergency that requires rapid detection and
intervention to improve survival rates and reduce long-term complications. Conventional
monitoring systems in Cardiac Intensive Care Units (CICUs) rely on threshold-based alarms and
manual assessments, which often fail to detect subtle physiological changes leading to cardiac
arrest. This study proposes an integrated machine learning and statistical model approach for the
early detection of cardiac arrest in newborns, leveraging real-time vital signs data, physiological
trends, and predictive analytics. The system utilizes supervised learning algorithms such as
Random Forest, Long Short-Term Memory (LSTM), and Support Vector Machines (SVM) in
combination with statistical anomaly detection techniques to identify high-risk patterns before a
cardiac event occurs. The model is trained on electronic health records (EHR) and real-time
patient monitoring data, enabling early warning alerts for healthcare professionals. Experimental
results demonstrate that the proposed approach significantly improves prediction accuracy,
sensitivity, and response time compared to traditional monitoring methods. This research
highlights the potential of AI-driven early warning systems to enhance neonatal cardiac care,
reduce mortality, and improve clinical decision-making in CICUs.