AI-based differential diagnosis and localization of idiopathic PVC using 12-lead ECG: multicenter registry study (EP-AI ASSIST)
C O Seo, S W Lee, Y J Kim, H S Lee, J H Jeong, Y G Kim, J M Shim, D H Kim, Y H Kim, Y S Baek, J I ChoiAbstract
Background
Pre-procedural identification of premature ventricular contraction (PVC) or ventricular tachycardia (VT) origins is crucial not only for determining the optimal treatment strategy but also for estimating procedural difficulty and predicting ablation success. Various 12-lead ECG-based algorithms predicting the origins rely heavily on operator expertise and exhibit limited reproducibility. Transformer-based deep learning (DL) models may provide an objective and reliable tool that support clinical decision-making and improves procedural planning for ablation.
Objectives
This study aimed to develop a transformer-based artificial intelligence (AI) model capable of predicting the origin of idiopathic PVC or VT using only standard 12-lead ECGs.
Methods
A total of 584 patients (mean age 46.5 ± 14.3 years; 255 men, 329 women) who underwent catheter ablation for idiopathic PVC or VT was included for this retrospective study. 470 standard 10-second 12-lead ECGs were selected after ECG quality review. Two transformer-based models were independently developed using an 7:1:2 split for training, validation, and testing datasets. Final localization was derived by combining output probabilities. Model performance was evaluated against ablation-confirmed the PVC or VT origins (Figure 1).
Results
The PVC or VT origins were classified as follows: RVOT (295, 50.5%), LVOT (93, 15.9%), anterolateral mitral annulus (37, 6.3%), posteromedial papillary muscle (21, 3.6%), RV inflow (16, 2.7%), LV summit (14, 2.4%), anterolateral papillary muscle (13, 2.2%), and others (95, 16.3%). The model achieved AUCs of 0.98 for LVOT, 0.91 for RVOT, 0.84 for LV Non-OT, and 0.93 for RV Non-OT, with a macro-average AUC of 0.915 (Figure 2). Performance remained consistent across internal validation, supporting its clinical utility.
Conclusion
This study presents a clinically validated transformer-based DL model capable of accurately localizing the origin of PVC or VT using standard 12-lead ECG. The model consistently achieved high diagnostic performance and may serve as a valuable non-invasive tool to support before ablation planning in patients with idiopathic PVC or VT.