Computer detection of atrioventricular dissociation from surface electrocardiograms during wide QRS complex tachycardias.

Abstract

Differentiation of wide QRS complex tachycardias on surface electrocardiograms is difficult for physicians and computers due in part to their inability to identify atrial activity, specifically atrioventricular (AV) dissociation. We studied 20 examples of AV associated rhythms and 17 examples of AV dissociated ventricular tachycardia. We applied an algorithm consisting of subtraction of a mean beat from each individual beat in leads II and V1 to generate remainder electrocardiograms. The remainder electrocardiograms were visually inspected for the presence of P wave candidates and then autocorrelated. AV dissociated P wave candidates were evident on visual inspection of remainder electrocardiograms in none of 20 AV associated and 15 of 17 AV dissociated rhythms. Atrial cycle length and the presence of AV dissociation were automatically detected by applying a peak selection algorithm to the autocorrelation function. AV association was detected in all 20 AV associated rhythms and AV dissociation was detected for 11 of 17 AV dissociated rhythms (sensitivity 65%, specificity 100%, positive and negative predictive accuracy 100%, 77%). The correlation coefficient of detected vs true atrial cycle length for the 11 correctly detected AV dissociated rhythms was r = .98. Visual inspection of the remainder electrocardiograms along with the original electrocardiogram may increase the ease with which human readers can identify the presence of AV dissociation and thus diagnose ventricular tachycardia. Computer diagnosis of wide QRS complex tachycardias should be significantly improved by use of this algorithm.