Multi-parametric algorithm for premature ventricular contractions detection and counting

A multi-parametric algorithm for the premature ventricular contractions detection based on both heart rhythm and ECG shape alternations is proposed. Six different QRS shape metrics are tested and their weighted combination is used to suggest an aggregated shape parameter. In the multi-lead ECG analysis each lead is weighted proportionally to its individual signal-to-noise ratio. Our results indicate that the proposed combined algorithm outperforms several complementary metrics in PVCs detection accuracy yielding ultimate performance characterized by Se=82.70, Sp=98.82 and +P=82.97 for the 2-lead MIT-BIH Arrhythmia Database as well as Se=92.87, Sp=99.64 and +P=97.03 for the St. Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database.