Artificial Intelligence Enabled Noninvasive Mapping of Cardiac Arrhythmia Origins Using the 12-Lead Electrocardiogram

Artificial intelligence (AI) is reshaping cardiac electrophysiology by extracting information from electrocardiograms that exceeds human visual interpretation. While most AI applications have focused on arrhythmia detection and risk stratification, a newer and more targeted use case is emerging: noninvasive localisation of arrhythmia origins from a standard 12-lead electrocardiogram (ECG). This review examines the current evidence supporting AI-driven ECG-based localisation of arrhythmogenic foci, with an emphasis on accessory pathways and idiopathic ventricular arrhythmias. Recent machine learning and deep learning models have demonstrated high accuracy in predicting anatomical sites of origin. These advances suggest that AI can capture subtle spatiotemporal electrical patterns that correlate with specific cardiac regions, offering a form of virtual pre-procedural mapping. Early clinical data indicate that AI-informed localisation may shorten procedure duration, reduce fluoroscopy exposure, and improve ablation efficiency without compromising safety. The review also discusses key technical innovations, including convolutional neural networks, multimodal data integration, and strategies to enhance model generalisability and interpretability. Important challenges remain, particularly around external validation, data quality, clinician trust, and workflow integration. Nevertheless, AI-driven ECG localisation represents a conceptual shift from diagnostic support toward therapeutic guidance in electrophysiology. As validation studies and translational research advance, this technology has the potential to transform pre-procedural planning and intraprocedural decision-making, thereby making arrhythmia ablation more precise, efficient, and accessible.