CNN-RNN-Bayesian Hybrid Method for Predicting Neonatal ICU Cardiac Arrests

Abstract

Infant cardiac arrest is a serious medical emergency that needs to be identified quickly in order to be effectively treated.  The goal of this study is to apply sophisticated statistical techniques to create a Cardiac Machine Learning Model (CMLM) that can predict neonatal cardiac arrest in the Cardiac Intensive Care Unit (CICU).  The model makes use of physiological markers and makes use of prediction methods like logistic regression and support vector machines.  The diagnostic procedure is enhanced by imaging techniques such as computed tomography and echocardiography.  With a delta-p value of 0.912, FDR of 0.894, FOR of 0.076, prevalence threshold of 0.859, and CSI of 0.842 in training and similar metrics in testing, the suggested CMLM showed excellent performance.  These findings point to the robustness and dependability of the model. The CMLM has the potential to dramatically lower neonatal mortality and morbidity rates by facilitating the early diagnosis of cardiac arrest episodes, which would improve outcomes for critically unwell infants in the intensive care unit.