Artificial Intelligence-Enabled Electrocardiography in Practice: A State-of-the-Art Review

Figures

• 

  Figure 1. Foundations of AI-ECG–driven digital biomarker development and clinical decision support. Schematic illustrating the core components required for the development of AI-enabled digital biomarkers: 1) high-dimensional input data (e.g., unstructured notes or imaging), 2) gold-standard labels, 3) sufficient training cases, and 4) high-performance computing resources. When successfully trained and validated, these AI models can transform ECG data into digital biomarkers that support diagnosis, prognosis, therapy selection, remote monitoring, physician efficiency, and multimodal data integration. These biomarkers ultimately contribute to enhanced clinical decision-making. AI = artificial intelligence; AI-ECG = artificial intelligence-enabled electrocardiography; ECG = electrocardiogram.

• 

  Figure 2. Future directions for AI-ECG integration in clinical and remote settings. Illustration of the dual trajectory shaping the future of AI-ECG. On the left, hospital-centered progress focuses on performance refinement, workflow optimization, and evidence-based validation. On the right, remote care advancements emphasize the development of smartwatch and portable ECG devices, secure and scalable data infrastructure, and early detection of clinical deterioration to prevent rehospitalization. Together, these directions represent a convergence of centralized and decentralized healthcare innovations driven by AI-ECG. AI-ECG = artificial intelligence-enabled electrocardiography; ECG = electrocardiogram.

• 

  Figure 3. Barriers and solutions for real-world integration of AI-ECG. Overview of 6 key challenges and corresponding solutions for achieving scalable, equitable, real-world integration of AI-ECG. These include: 1) lack of economic justification, addressed through cost-effectiveness studies; 2) limited clinician familiarity, resolved via structured AI education; 3) inconsistent data quality and demographic bias, mitigated by standardization and diverse datasets; 4) static regulatory frameworks, requiring adaptive validation and governance; 5) limited model transparency, countered with explainable AI tools; and 6) suboptimal IT infrastructure, requiring workflow-integrated system design. These coordinated efforts are essential to transition AI-ECG from research to routine clinical practice. AI = artificial intelligence; AI-ECG = artificial intelligence-enabled electrocardiography; ECG = electrocardiogram; EHR = electronic health record; XAI = explainable artificial intelligence.