A Long Short-Term Memory Network to Classify Myocardial Infarction Using Vectorcardiographic Ventricular Depolarization and Repolarization

Abstract

QT interval beat-to-beat variability has indicated diag-nostic/prognostic abilities in myocardial infarction. Furthermore, research has suggested that vectorcardiography has superior diagnostic abilities compared to the standard electrocardiogram in myocardial infarction. This study aimed to assess the ability of vectorcardiographic ventricular depolarization and repolarization to classify myocardial infarction patients versus control subjects. 147 vectorcardiogram recordings (78 MI vs. 69 Control) were obtained from the PTB database. For each recording, 60 QRS-complex and T-wave VCG beats were extracted using the Two-Dimensional Signal Warping algorithm. An inhomogeneous three-dimensional template adaptation scheme was applied on each QRS-loop and T-loop to capture subtle morphological changes from beat-to-beat. Training was performed on a regularized three-layer long short-term memory network. The classifier produced test set classification results with an overall 89.1% accuracy, 89.1% sensitivity and 90.0% specificity. In conclusion, high classification accuracy has been achieved on a relatively small subset of the PTB database. Future work will look to improve the classification results by extending the analysis across the entire PTB database.