Sensitivity and specificity of a dual-chamber arrhythmia recognition algorithm for implantable devices

Abstract

Present ventricular rate-based arrhythmia detection algorithms lack specificity. Using a training set of 109 endocardial electrogram recordings, a sensitive and specific dual-chamber arrhythmia recognition algorithm has been developed. The algorithm uses atrial and ventricular rates, irregularity, degree of beat-to-beat similarity, and measure of electrogram complex distinctiveness to arrive at a diagnostic conclusion. A test set of 121 endocardial electrogram recordings obtained during provocative electrophysiology studies was then used for blinded validation of the algorithm. In normal rhythm, 1:1 tachycardia, atrial tachycardia, atrial flutter, atrial fibrillation, ventricular tachycardia, and ventricular fibrillation, the percentages of sensitivity/specificity were, respectively, 100/99, 100/99, 80/99, 89/98, 91/97, 92/100, and 100/98. Although ventricular rate alone can usually distinguish normal rhythm, ventricular tachycardia, and ventricular fibrillation, it is confounded by atrial arrhythmias and 1:1 tachycardias. When tested on a database, a ventricular rate-only algorithm resulted in sensitivity/specificity of 100/65, 90/78, and 100/99%, respectively, for these three rhythms. Therefore, the dual-chamber algorithm based on both temporal and morphologic measures provides better distinction of normal rhythm and ventricular tachycardia than existing methods, without sacrificing sensitivity