In vivo validation of a novel algorithm for automatic premature ventricular contractions recognition

Template-matching algorithms are routinely used in the catheter ablation of patients with premature ventricular contractions (PVCs). However, systematic analysis of the accuracy and spatial resolution of such systems is lacking. Therefore, the aim of this evaluation was to perform a systematic in vivo validation of performance of a novel automated template-matching algorithm.

In a porcine model, paced beats simulating PVCs from different origins were investigated. The ability to discriminate between sinus rhythm and PVCs was tested by simulating PVCs using sequential pacing from different cardiac chambers. The accuracy of the algorithm in correctly classifying PVCs was reviewed by an independent investigator. In addition, the spatial resolution of pace matching was evaluated by assessing the QRS morphology discrimination at a distance of 0, 2, 4, and 6 mm of a simulated PVCs focus. The specificity of the algorithm for recognizing simulated PVCs was 99.6% and the sensitivity was 85.3%. There was a significant difference in the discrimination metric discrimination metric (with 0% being a perfect match and 100% being no correlation) between PVC origin (median 0%, interquartile range (IQR) 0-2%) versus at 2 mm (5%, IQR 2-7%), 4 mm (16%, IQR 11-21%), and 6 mm (24%, IQR 19-28%, P < 0.001 for all). The c-statistic for discrimination between PVC origin and a distance ≥ 2 mm was 0.93.

Automated template matching had high specificity and sensitivity, with good spatial discrimination and a pace-mapping resolution in range of 2 mm. Clinical application of this algorithm may assist in the interventional treatment of patients with PVCs.