Automated detection and mapping of electrical activation when electrogram morphology is complex

BackgroundMapping of cardiac electrical activity can be difficult when electrogram morphology is complex. Complex morphology (multiple and changing deflections) causes activation maps to vary when constructed by different analysts, particularly at areas with spatially varying conduction pattern. An algorithm was developed to automatically detect electrogram activation time which is robust to complex morphology.MethodElectrograms, many of which were complex, were collected from 320 canine epicardial border zone sites in five experiments. A library of electrogram activation times were manually marked a priori by two expert analysts. Then an algorithm which combined correlation and error functions was used to compare each input electrogram to library electrogram patterns. The closest match of input to library electrogram was used to estimate activation time. Once activation times at 320 sites were determined, activation maps were automatically constructed on a computerized grid. The algorithm was validated by comparison with activation times determined by the analysts.ResultsThe mean difference between manual and automated marking of activation time in electrograms acquired during reentrant ventricular tachycardia was 2.1 ± 3.9 ms. The mean sensitivity and positive predictive value were 95.9% and 83.8% respectively. The computer-automated marking process was completed within a few seconds and was robust to fractionated electrograms. Measurement error was mostly attributable to 60 Hz noise, which can be rectified with filtering.ConclusionsThe automated algorithm is useful for rapid and accurate automatic marking of multichannel electrograms, some of which may be fractionated, as well as for real-time display of activation maps in clinical electrophysiology or research studies.

► A robust signal processing method is described for heart arrhythmias. ► The algorithm accurately marks complex electrograms for activation. ► The method uses two quantitative marking criteria to minimize errors. ► The activation marking and mapping algorithm is automated. ► In test data, mean error in marking was 2.1 ± 3.9 ms.