A machine learning approach for classifying healthy and infarcted patients using heart rate variabilities derived vector magnitude

Abstract

According to the World Health Organization, Heart disease is the number one killer of humans, with coronary heart disease (CHD) being the most common type of heart disease. CHD leads to myocardial ischemia (MI) or infarction. Several clinical tests are available to assist physicians in diagnosing MI or infarcted (unhealthy) patients. However, diagnostic tests can be costly, invasive, and unreliable in identifying patients with declining coronary health conditions. This study investigated the application of Machine Learning (ML) techniques on the Vector Magnitude (VM) data of heart signals generated via Vectorcardiography (VCG) to classify unhealthy patients from healthy patients. Patients with MI, a CHD, are identified as ill patients. Three machine-learning classification techniques: Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Decision Trees (DT), were applied to classify healthy and unhealthy (MI) patients. The heart signal dataset was acquired from the Physikalisch-Technische Bundesanstalt (PTB) Diagnostic electrocardiogram (ECG) Database. A 10-fold cross-validation sampling method was used to improve the predictability of the sample. Results from ML techniques produced high classification sensitivity, specificity, and accuracy. ML analysis findings indicated that DT is the best predictor for classification accuracy, followed by SVM and ANN. The future study goal is to expand the study with forward-looking data and the right sample size for clinical validity and support the high accuracy results.

Keywords

• Coronary disease
• Heartrate variability
• Classification
• k-fold cross-validation