Artificial Intelligence in Arrhythmology

This doctoral thesis explores the integration of artificial intelligence (AI) into the field of clinical arrhythmology and electrophysiology, emphasizing its transformative potential in arrhythmia management. The research is structured around three studies that collectively demonstrate AI’s capabilities in prediction, localization, and diagnostic analysis within this broad subspecialty field of cardiology. The first study introduces AFA-Recur, a machine learning web calculator developed from the ESC EORP AFA-LT registry. This tool predicts atrial fibrillation recurrence post-ablation by analyzing clinical variables, highglighting the potential of AI in personalized patient risk stratification. The second study presents a machinelearning model that integrates electrocardiograms (ECGs) and clinical data to automatically and interpretably predict the site of origin in outflow tract ventricular arrhythmias. This approach enhances the precision of arrhythmia localization, facilitating targeted therapeutic interventions. The third study evaluates the efficacy of convolutional neural networks (CNNs) in ECG analysis, comparing standard twelve-lead and single-lead setups. The findings suggest that CNN-enabled analysis can maintain diagnostic accuracy even with reduced lead configurations, indicating potential for scalable and accessible arrhythmia detection. Collectively, these studies underscore AI’s emerging role in clinical arrhyhtmology and electrophysiology. The integration of AI into clinical practice holds promise for advancing patient care in this field.