ADVANCED PREDICTIVE ANALYTICS FOR EARLY CARDIAC ARREST DETECTION IN NEWBORN INFANTS

Abstract

A medical emergency that is both common and terrifying is unexpected cardiac arrest in infants. The most effective care and treatment for these infants can only be administered if they are discovered at an early stage. In recent years, scientists have been conducting research to identify potential biomarkers and symptoms of neonatal cardiac arrest in order to create more precise and efficient diagnostic tools for early detection. Echocardiography and computed tomography are two of the numerous imaging modalities that have the potential to identify cardiac arrest at an early stage. The objective of this research is to rapidly identify neonatal cardiac arrest in the CICU by constructing a cardiac machine learning model (CMLM) that employs statistical models. The integration of the neonate's physiological data enabled us to ascertain the frequency of cardiac arrests. Predictive models for cardiac arrest were developed using two statistical modeling techniques: logistic regression and support vector machines. The proposed procedure will be implemented by the CICU to expedite the detection of neonatal cardiac arrest. In the training (Tr) comparative zone, the proposed CMLA achieved a 0.912 delta-p, 0.894 FDR, 0.076 FOR, 0.859 prevalence threshold, and 0.842 CSI. The CMLA that was recommended in the testing (Ts) comparison zone had the following values: 0.896 delta-p, 0.878 FDR, 0.061 FOR, 0.844 prevalence threshold, and 0.827 CSI. As an outcome, neonatal cardiac arrest-related mortality and morbidity will be diminished.