AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Structural heart disease (SHD) is often underdiagnosed, hindering timely treatment.
• Echocardiography is crucial for diagnosis but faces limitations in screening capacity and efficiency.
• Artificial intelligence-enabled 12-lead electrocardiography (AI-ECG) offers a potential solution for pre-echocardiographic triage.

Purpose of the Study:

• To review the utility of AI-ECG as a pre-echocardiographic triage tool for various SHDs.
• To evaluate AI-ECG's performance in different intended-use orientations: safety-net screening and gatekeeper triage.
• To synthesize current evidence and identify areas for future research and implementation.

Main Methods:

• Systematic review and synthesis of evidence on AI-ECG for SHD detection.
• Evaluation of AI-ECG models for specific conditions: reduced ejection fraction, valvular disease, hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary hypertension.
• Analysis of AI-ECG's role in safety-net and gatekeeper triage strategies.

Main Results:

• AI-ECG shows strongest evidence for detecting reduced left ventricular ejection fraction (LVEF), with available implementation and economic data.
• Models for valvular and composite SHD show promise for improving referral accuracy.
• Applications for hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary hypertension are less mature.
• AI-ECG is best positioned as an additive safety-net tool for enhancing case finding.

Conclusions:

• AI-ECG can serve as a valuable safety-net tool to identify patients needing echocardiography earlier.
• Gatekeeper triage use of AI-ECG to defer echocardiography requires further prospective validation and safeguards.
• Implementation considerations include false-positive interpretation, workflow integration, equity, and regulation.