Linear and Nonlinear Features for Myocardial Infarction Detection Using Support Vector Machine on 12-Lead ECG Recordings

The development of non-invasive techniques to assess cardiovascular risks has grown rapidly. In this sense, a multi-lead electrocardiogram (ECG) provides useful information to diagnose myocardial infarction (MI), the leading cause of death worldwide. In this paper we used a support vector machine (SVM) to detect MI by exploiting temporal, morphological and nonlinear features extracted from 12-lead ECG recording from the PTB Diagnostic ECG database. Temporal features correspond to QT, ST-T and RR intervals, morphological features were extracted from P and T waves, and QRS complexes, and nonlinear features correspond to the sample entropy of QT, ST-T and RR intervals. A 10-fold Monte Carlo cross-validation was implemented by randomly splitting the data set into training (70%) and test (30%) sets with balanced classes. Sensitivity of 97.33%, specificity of 96.67%, and accuracy of 97.00% were obtained by jointly exploiting temporal, morphological and nonlinear features by the SVM. The inclusion of entropy favors the detection of healthy control cases because the information of signal regularity improves the specificity of classification.