A Machine Learning and Statistical Modeling Framework for Early Detection of Neonatal Cardiac Arrest

Abstract

In order to preserve the newborn's life and enhance their health, neonatal cardiac arrest must be promptly diagnosed and treated. Early warning indicators are not adequately addressed by conventional monitoring methods in neonatal intensive care units (NICUs). In order to promptly identify neonates in cardiac arrest, this investigation suggests the utilization of physiological and clinical indicators, including respiration, heart rate, oxygen saturation, birth weight, and gestational age. The technology is dependent on machine learning methods and statistical models. In order to enhance the accuracy of predictions, numerous machine learning techniques are currently being investigated. Naïve Bayes, ANN, Decision Tree Classifier, SVM, and Logistic Regression models comprise the ensemble. Statistical preprocessing and feature extraction enhance the efficacy of models and minimize the number of inaccurate predictions. It was determined by a paper that the prediction system is capable of identifying neonates who are at risk of cardiac arrest. Additionally, it is precise and dependable. The proposed method would enhance the survival rate of infants and enable early intervention, which would be advantageous to neonatal intensive care unit physicians.