Artificial intelligence-enabled electrocardiography from scientific research to clinical application

Abstract Recent advancements in artificial intelligence (AI) have revolutionized the application of electrocardiography (ECG) in cardiovascular diagnostics. This review highlights the transformative impact of AI on traditional ECG analysis, detailing how deep learning algorithms are overcoming the limitations of human interpretation and conventional diagnostic criteria. Historically, ECG interpretation has relied on well-established, physiologically-based criteria. The advancement of AI-ECG is marked by its capacity to process complex high-dimensional data directly from raw signals, revealing patterns often missed by conventional methods. Notably, AI models have identified signs of asymptomatic low ejection fraction and paroxysmal atrial fibrillation during normal sinus rhythm, enabling earlier clinical intervention. In addition to improved diagnostic utility, AI-ECG offers promising applications in risk stratification and community screening. Several randomized controlled trials (RCTs) have shown that integrating AI into clinical workflows not only reduces critical intervention times but also identifies patients at elevated risk of adverse outcomes. Future directions involve integrating additional clinical data sources, improving model interpretability through explainable AI, and developing unified platforms to manage outputs from multiple models.