A Machine Learning Approach Using Statistical Models for Early Detection of Cardiac Arrest in Newborn Babies in the Cardiac Intensive Care Unit

Abstract

Introduction: Caring for newborn babies is one of the major concerns for doctors, and taking care of his/her life is very challenging and sometimes critical. Cardiac attack is one of the severe issues that requires immediate medical support for the newborn baby. Providing some shocks using a device is one of the immediate treatments to save the lives of babies. The computation and prediction of cardiovascular status is one of the challenging tasks. In this work, for the early detection of cardiac arrest in newborn babies used machine learning approach with statistical method Decision Tree Classifier (DTC), compared with other methods (ANN, SVM, LR) for the cardiac prior heart failure chances at the different levels. The target rate of cardiac arrest detected is 35% and the level of no cardiac arrest found is 65% which is achieved with the proposed methodology, and this prediction level helps to recognize the level of cardiac attack for newborn children. Also, analyse the different parameters like false rate, false observation, stability rate, etc., that will help for the proper analysis of the collected data with a stability rate of 1.6700% and precision rate of 60.0390%. These results help the healthcare professional to handle the different levels of heart failure in newborns.

Objectives: Early detection of cardiac arrest in newborn babies using machine learning approach with statistical method, perform comparative analysis.

Methods: Machine Learning approach with statistical method Decision Tree Classifier (DTC), compared with other methods (ANN, SVM, LR) for the cardiac prior heart failure chances at the different levels.

Results: Models once used rely on ML that allows for analysis and determining diseases related to the heart using insights from several superior ML strategies. Preprocessing is one of the important techniques in ML strategies that will discriminate the cardiac arrest at an early neonatal time point. And, as a result of this phase, the full data is now ready for analysis by the machine learning process. This process ensures that the data collected is suitable for the machine learning algorithms used. the data set would have been divided into a validation, training, and testing set so that models could be fairly compared and evaluated to get the best performance. In early detection, the feature extraction method could identify cardiac arrest for infants, and it is one of the methods by which important features could be extracted from the data to detect cardiac arrest.

Conclusions: The proposed Decision Tree Classifier of the Machine Learning model has important implications for the previous identification of cardiac infarctions in neonates. It is critical to provide appropriate therapy for babies in the ICU, which will allow accurate identification of babies experiencing severe cardiac arrest or at high risk thereof. Overall, the results obtained would assist the health professional to make informed identifications of cardiac failure, which may range through various levels of diagnosis.