Master of Science
AARON L ROBINSON
Sudden cardiac arrest is a deadly illness. Cardiac arrest often occurs without previews symptoms which makes it dangerous and hard to predict. Sudden cardiac arrest happens due to a heart’s electrical system problem. This causes the heart to stop pumping blood and as a consequence, the flow of blood in the body ceases. The percentage of death from cardiac arrest is almost 75% in the hospital. Therefore, the ability to predict plausible cardiac arrest from the statistics, collected data, and heart-related analysis can provide a critical benefit to health care professionals. This project intends to study the currently available data on sudden cardiac arrest and develop an algorithm to have accuracy above 90%. The cardiac arrest prediction model intends to predict sudden cardiac arrest based on vital signs and lab test results. Those vital signs are blood pressure, respiration, body temperature, and heart rate. The lab tests included oxygen level in the blood, potassium, glucose, alkalinity of the blood, chloride, carbon dioxide, magnesium, hemoglobin, nitrogen in the blood, measuring the time for blood to clot, and phosphorous level. This research computes eleven machine learning classifiers: Logistic Regression, Random Forest, Decision Trees, Nearest Neighbors, Linear SVM, RBF SVM, Neural Net, AdaBoost, Gaussian Process, Naive Bayes, and QDA. AdaBoost is the classifier with the highest performance. This project proposes an alternative method for cardiac arrest prediction using vital signs and lab test findings using machine learning.
Dissertation or thesis originally submitted to ProQuest
Torres, Jose Manuel, "Predicting Sudden Cardiac Arrest Using Machine Learning From ICU Patients Records" (2022). Electronic Theses and Dissertations. 3214.