[Wolff-Parkinson-White syndrome : Comparison of different algorithms]

Wolff-Parkinson-White (WPW) syndrome is characterised by accessory pathways that bypass the normal atrioventricular conduction system. Precise preprocedural localisation is pivotal for optimising ablation strategy, minimising complications, and reducing radiation exposure.

This review systematically analyses ECG-based algorithms for accessory pathway localisation, including classical and recent rule-based approaches, as well as modern deep learning models.

Classical algorithms showed variable accuracy ranging from 72% (Milstein) to 92% (D'Avila, St. George). Modern rule-based algorithms demonstrate significantly improved performance: EASY-WPW achieved 93% accuracy (sensitivity 92%, specificity 99%), and SMART-WPW reached 97% (sensitivity 96%, specificity 100%) using a 12-location clock-face model. DL approaches achieved an 84% accuracy with AUROC 0.92, significantly outperforming classical algorithms (Milstein AUROC 0.81, Arruda AUROC 0.80). The DL model enables automatic analysis, reduces interobserver variability, and identifies parahisian pathways and locations requiring transseptal puncture. Both EASY-WPW and SMART-WPW showed excellent results in pediatric populations.

Both validated ECG algorithms and deep learning models represent valuable tools for preinterventional planning in patients with WPW syndrome. Modern rule-based algorithms offer excellent diagnostic accuracy with sensitivities and specificities exceeding 90%. The integration of artificial intelligence (AI) and multimodal approaches promises further improvements in accessory pathway localisation.