A Hybrid Deep Learning and Statistical Framework for Cardiac Arrest Detection in Neonatal Intensive Care Units

Neonatal cardiac arrest, particularly in preterm infants admitted to Neonatal Intensive Care Units (NICUs), is a serious condition with a need for early and accurate detection. This study proposes a hybrid framework combining deep learning models with statistical techniques that can improve the reliability and accuracy of predictions for the occurrence of cardiac arrest in neonates. Using the Preterm Infant Cardio- Respiratory Signals Database, we extract spatial features from ECG and respiratory signals using Convolutional Neural Networks (CNNs) and capture temporal dependencies using Long Short-Term Memory (LSTM) networks. The framework is complemented by statistical models, including Support Vector Machines (SVM) and Random Forests, to enhance classification robustness and interpretability. The proposed system preprocesses raw physiological signals to remove noise, extracts meaningful features, and generates real-time predictions with high Accuracy of 92. Comparative analysis with existing methods demonstrates significant improvements in performance, highlighting the system's potential for early detection and real-time intervention in NICUs. This hybrid approach not only improves neonatal survival rates but also ensures interpretability, enabling healthcare providers to make confident and informed decisions.