In-silico pace mapping identifies pacing sites more accurately than inverse body surface potential mapping

Background: Electrocardiographic imaging (ECGi) is a non-invasive technique for ventricular tachycardia (VT) ablation planning. However, it is limited to reconstructing epicardial surface activation. In-silico pace mapping combines a personalized computational model with clinical electrocardiograms (ECGs) to generate a virtual 3D pace map.

Objective: To compare the ability of ECGi and in-silico pace mapping to determine the site of ventricular pacing.

Methods: ECGi recordings were collected during left ventricle (LV) (endocardial: N=5, epicardial: N=1), septal (N=3) and right ventricle (RV) apical pacing (N=15) along with computed tomography (CT). Personalized CT-based ventricular-torso computational models were created and aligned with the 252 ECGi vest electrodes. Ventricles were paced at 1000 random sites, and corresponding body surface potentials (BSPs) and ECGs were computed. In-silico pace maps were then reconstructed by correlating all simulated ECGs or BSPs with the corresponding paced clinical signals. The distance (d) between the pacing electrode (ground truth) and the location with the strongest correlation was determined; for ECGi, the site with earliest activation time was used.

Results: In-silico pace mapping consistently outperformed ECGi in locating the pacing origin, with the best results when all BSPs were used. During LV pacing, spatial accuracy of in-silico pacing mapping was 9.5mm with BSPs and 12.2mm when using ECGs, compared to 30.8mm when using ECGi. During RV pacing, d = 26.1mm (BSPs), 30.9mm (ECGs) and 29.1mm (ECGi).

Conclusion: In-silico pace mapping is more accurate than ECGi in detecting paced activation. Performance was optimal when all BSPs were used and reduced during RV apical pacing.