Wolff Parkinson White (WPW) syndrome, a significant cause of ventricular pre-excitation, presents diagnostic challenges in patients where early detection is crucial to prevent life threatening arrhythmias. The condition is characterized by electrocardiogram (ECG) features such as a short PR interval and a delta wave. Although traditional rule-based interpretation algorithms and feature-based machine learning (ML) models exist, they often lack the accuracy required for consistent and reliable WPW diagnosis. This study evaluates an end-to-end deep learning (DL) model for automated WPW detection from 10-second 12-lead ECGs in patients under 17 years old, a presentation that is relatively rare, compared to a ML model. Our results show that the DL model substantially outperforms the feature-based ML model, demonstrating superior accuracy in identifying pediatric WPW. To address the "black box" nature of DL, we applied Shapley additive explanations (SHAP) to improve model transparency. SHAP analysis visually highlights critical ECG segments, such as the delta wave, that most influence the model's decisions. This explainable AI-powered DL model provides clinicians with a powerful, transparent tool for precise WPW diagnosis, supporting improved clinical risk stratification and management.