New algorithms for QT dispersion analysis

In this paper, new algorithms for QT dispersion analysis on standard 12-lead ECG are described. A singular-value-decomposition (SVD)-based filter was designed to smooth T wave and a pattern recognition method was developed to find different T wave patterns and their peaks. Then a least-square curve fitting method was used to locate T end. A fuzzy thresholding approach was adopted to decide the final T peak and T end Subsequently, Q to T end dispersion and Q to T peak dispersion of all 12 leads and 6 precordial leads were computed. We also studied a SVD-based template matching far QT dispersion. We have tested the new QT dispersion algorithms on ECGs from normals, acute myocardial infarction patients, and serial ECGs for reproducibility. Results show that the new method is effective and consistent in differentiating normal subjects from abnormal group. The QT dispersion difference for reproducibility is within 1 to 2 sample points.