A reconstructed phase space approach for distinguishing ischemic from non-ischemic ST changes using Holter ECG data

A method for the classification of ST events through the use of reconstructed phase spaces of the ECG signal is proposed. There is a clinical need for the creation of an automated system for classification of ST events as ischemic or non-ischemic as existing ischemia detection methods are expensive, invasive, or both. The algorithm proposed herein attempts to classify events using the 16 beats surrounding a given ST event. The ST segment and T wave of each of these beats is embedded in a phase space and then modelled and classified through the use of Gaussian mixture models (GMM). Using ten-fold cross validation of available training data the sensitivity and specificity were 81.0% and 88.1% respectively.