Background
Mathematical modeling–based noninvasive mapping technologies are increasingly applied in cardiac electrophysiology to reconstruct ventricular activation patterns from standard surface electrocardiograms (ECGs), without the need for invasive cardiac catheterization. A novel 12-lead ECG-based noninvasive mapping algorithm employs an advanced inverse-solution framework integrating biophysical modeling and numerical optimization to identify arrhythmia origins with high spatial precision. This study presents independent clinical validation data, with a particular focus on inter-center reproducibility and real-world performance across diverse ventricular arrhythmias.
Methods
Fifteen ECG recordings from patients with premature ventricular contractions (PVC) undergoing catheter ablation were evaluated. Eleven cases met predefined quality and completeness criteria (4 cases were excluded due to incomplete data). Noninvasive activation maps were reconstructed using the noninvasive mapping algorithm and compared with invasive electroanatomic maps acquired using a commercially available mapping system. Two ordinal scoring metrics were applied: earliest-activation match (0–5), reflecting spatial correspondence of the earliest activation site, and activation-pattern agreement (0–5), assessing topographic similarity of the overall activation sequence. Representative examples are shown in Figure 1. Additionally, the Euclidean distance between earliest noninvasive and invasive activation points (mm) was measured (Figure 2). A mapping result was considered clinically acceptable when both scores were ≥3.
Results
Invasive mapping localized PVC origins to the right ventricular (RV) outflow tract in 4 cases (36.4%), left ventricular (LV) outflow tract in 2 cases (18.2%), lateral basal RV in 1 case (9.1%), and inferior or anterior LV in 3 cases (27.3%).
In 5 cases (45.5%), the earliest noninvasive activation site was localized to the exact corresponding myocardial segment with concordant activation patterns, 3 cases (27.3%) demonstrated localization to immediately adjacent or overlapping anatomical regions, and 3 cases (27.3%) exhibited activation mapping to neighboring segments with preservation of accurate propagation patterns. The mean localization deviation between noninvasive and invasive earliest-activation points was 12±7 mm.
Conclusions
This independent clinical validation demonstrates reproducibility and spatial accuracy of a novel 12-lead–based noninvasive mapping algorithm. High degrees of concordance were observed between noninvasive and invasive activation mapping across a broad spectrum of PVC morphologies.
Figure 1
Figure 2
