A Neural Network model achieved 94% accuracy in the early detection of cardiac arrest in newborns, outperforming other evaluated machine learning algorithms.
Does the Cardiac Machine Learning Model (CMLM) accurately predict the probability of cardiac arrest in neonates in the CICU?
A machine learning model using physiological parameters demonstrates strong predictive metrics for the early detection of cardiac arrest in high-risk neonates.
Absolute Event Rate: 94% vs 92%
Newborn cardiac arrest is a serious medical emergency that needs to be recognised early to improve survival and decrease complications. In this work, we propose an approach using statistical models based on machine learning for early detection of cardiac arrest in neonates in the Cardiac Intensive Care Unit (CICU). The model uses physiological parameters of newborns and uses statistical methods such as logistic regression and support vector machines to predict the probability of cardiac arrest. The proposed Cardiac Machine Learning Model (CMLM) has shown promising results with delta-p value of 0.912, false discovery rate (FDR) of 0.894, false omission rate (FOR) of 0.076, prevalence threshold of 0.859, and critical success index (CSI) of 0.842 during training. In tests, the model produces comparable results, with a delta-p of 0.896, FDR of 0.878, FOR of 0.061, prevalence threshold of 0.844 and CSI of 0.827. The system helps in early detection of high-risk newborns, thus enabling timely medical intervention and significantly reducing mortality and morbidity rates in neonatal care.
Johar et al. (Fri,) conducted a other in Neonatal cardiac arrest. Neural Network model vs. Other machine learning models (Support Vector Machine, Logistic Regression, Naive Bayes) was evaluated on Prediction accuracy. A Neural Network model achieved 94% accuracy in the early detection of cardiac arrest in newborns, outperforming other evaluated machine learning algorithms.