A comprehensive review of artificial intelligence in Electrocardiogram diagnostics: Integrating knowledge map and meta-analysis approaches

Electrocardiogram (ECG) record the heart’s electrical activity and is vital for continuous cardiac monitoring. While dozens of reviews have surveyed deep learning approaches to ECG analysis, they rarely address the field’s full scope. Here, we systematically review 2,990 ECG studies published over the past decade and perform a meta-analysis on 58 articles evaluating algorithmic performance for atrial fibrillation (AF), myocardial infarction (MI), and coronary artery disease (CAD). A literature-based knowledge map highlights machine learning and deep learning as dominant research trends. Our meta-analysis reveals that convolutional neural networks (CNNs) deliver the highest diagnostic accuracy for AF, MI, and CAD, though efficacy diminishes across those conditions. We also explore emerging methods, including large language models, and conclude by discussing outstanding challenges and future directions in data quality and diversity, model generalizability, clinical integration, and novel technology adoption. • Refined insights into ECG diagnostics leveraging knowledge mapping and meta-analysis. • Meta-analysis applied to consolidate ML prediction outcomes in ECG diagnostics. • Solutions for emerging challenges include multimodal data fusion and privacy-preserving tools. • Proposed personalized modeling and ethical frameworks to ensure fair AI-driven healthcare access.