Evaluation of a novel pacemaker-based electrogram morphology monitor algorithm in conduction system pacing
P Vijayaraman, W Zhang, L He, G Zeng, R Li, T Shi, T Brandon, J Cerkvenik, X Zhou, J CaoAbstract
Introduction
The need for 12-lead ECGs during follow-up of pacemaker patients with conduction system pacing (CSP) presents a clinical workflow challenge. CSP morphology monitor is a new algorithm designed to automatically monitor ventricular electrogram (EGM) morphology, aiding clinicians in confirming and maintaining effective CSP therapy.
Purpose
To evaluate the performance of the CSP morphology monitor algorithm in a clinical test dataset, with specificity and negative predictive value (NPV) as the primary end point.
Methods
To assess CSP capture, changes in far-field EGM (RV Tip-to-Can) were calculated using two metrics (Figure 1): Wavelet Match Score (0–100%), which quantifies similarity between a stored in-office template beat (e.g., at 5 V) and subsequent test beats using a Wavelet transform; and EGM-derived Left Ventricular Activation Time (LVAT), measured from the ventricular pacing marker to the center of the R-wave area, correlating with ECG LVAT. Prospective data were collected in pacemaker patients with left bundle branch area pacing using the Model 3830 lead. Device EGMs and 12-lead ECGs were acquired during pacing threshold tests at follow-up (unipolar/bipolar pacing; supine/sitting). Performance was assessed by comparing beat-by-beat algorithm outputs with expert adjudication using generalized estimating equations (thresholds: 70% Match Score, ±12% EGM LVAT Ratio).
Results
A total of 38 patients were included (mean age: 70 years, range 27–89; 45% male; follow-up visits 1-9 months post-implant). EGM templates and EGM LVAT during non-selective left bundle branch pacing (NS-LBBP) at 5 V were successfully created in all patients. The algorithm was evaluated with 11,346 observations for Wavelet Match Score and 11,331 for EGM LVAT. Changes in EGM morphology mainly included transitions from NS-LBBP to left ventricular septal pacing (LVSP) or selective LBBP (S-LBBP). Beats with loss of pacing capture were excluded. The algorithm demonstrated a specificity of 92.8% (95% confidence interval [CI]: 89.1–95.3%), and NPV of 98.5% (95% CI: 96.9–99.3%) for Wavelet Match Score; and a specificity of 97.1% (95% CI: 93.3–98.8%) and NPV of 97.2% (95% CI: 93.2–98.9%) for EGM LVAT in detecting changes in CSP capture. Of the 12 patients with transitions from NS-LBBP to LVSP, 10 (83%) were correctly identified using either the Wavelet Match Score or the EGM LVAT algorithm. The 2 patients who transitioned from NS-LBBP to S-LBBP were detected by the Wavelet Match Score algorithm, while EGM LVAT changes remained within 12%.
Conclusion
The CSP morphology monitor showed high specificity and negative predictive value for both Wavelet Match Score and EGM-derived LVAT. When Match Score is ≥ 70% or LVAT changes remain within 12%, a change in NS-LBBP capture is unlikely. This algorithm can detect clinically relevant EGM morphology changes and could help reduce unnecessary in-office follow-up visits for future pacemaker patients.