Computer Identification of Premature Ventricular Contractions in Electrocardiograms.

摘要

Algorithms to locate and identify premature ventricular contractions (PVC's) in single lead electrocardiograms are developed and evaluated on twelve 100 second electrocardiogram recordings. The location algorithms detect waveforms which are either QRS complexes or PVC's. The identification algorithms are then used to classify the waveforms as QRS complexes, PVC's, or premature beats. QRS and PVC classification is based on the evaluation of linear combinations of low sequency Walsh transform coefficients for each located waveform. Premature beat classification is based on a rhythm analysis. Of 1397 QRS complexes evaluated, 1386 or 99.21% were classified correctly.