Automatic delineation of single-lead electrocardiograph fiducial points based on the hierarchical triple-extreme-points model

Abstract

This paper introduces a novel single-lead electrocardiograph (ECG) automatic delineator based on the hierarchical triple-extreme-points model (HTM) and self-learning, featuring high robustness, low computational cost and mathematical simplicity. By applying HTM to obtaining the local morphological features of ECG signals, this method is capable of precisely detecting QRS complexes, P-wave and T-wave. In the experimental studies, this method was validated by several standard real-world ECG databases, including MIT-BIH arrhythmia, QT, European ST-T and TWA Challenge 2008 databases. For QRS detection, the average sensitivity value was 99.74% and the positive predictivity value was 99.80% for all databases. In the meantime, for P-wave and T-wave detection on QT, the sensitivity values were 99.49% and 99.81% respectively and the positive predictivity values were 98.82 and 99.80% respectively. As to delineation, the average maximum delineation error was no more than 4 ms and the standard deviation error was around 10ms for P-wave, QRS complex and T-wave.