Artificial Intelligence-Enabled ECG for Diastolic Dysfunction in Congenital Heart Disease: A Novel Risk...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Assessing diastolic function in adult congenital heart disease (ACHD) presents challenges due to complex patient anatomy and varied physiology.
• Echocardiography is the standard but can be limited in ACHD patients.

Purpose of the Study:

• To evaluate an artificial intelligence-enabled electrocardiogram (AI-ECG) model for grading diastolic dysfunction in ACHD patients.
• To assess the correlation of AI-ECG grading with echocardiographic, invasive hemodynamic data, and clinical outcomes.

Main Methods:

• A retrospective analysis of 6,741 ACHD patients from the Mayo Clinic ACHD Registry (2000-2023).
• A validated deep neural network was used to assign an AI-ECG diastolic grade (0-3) from the earliest ECG.
• Associations with echocardiography, hemodynamics, and mortality were evaluated using various statistical methods.

Main Results:

• Higher AI-ECG diastolic grades correlated with increased age, CHD complexity, comorbidities, and elevated N-terminal pro-B-type natriuretic peptide levels.
• AI-ECG pressure estimates showed correlation with echocardiographic strain measures.
• Invasive hemodynamics, including right atrial and wedge pressures, increased with higher AI-ECG grades.
• Diastolic grades 2 and 3 independently predicted mortality (HR 1.38-1.63).
• The AI-ECG model demonstrated good discrimination for elevated pulmonary artery wedge pressure (AUC 0.74).

Conclusions:

• AI-ECG diastolic grading shows strong correlation with established measures of cardiac filling pressures in ACHD.
• The AI-ECG model serves as an independent predictor of mortality in this population.
• AI-ECG grading is a promising, scalable, and noninvasive tool for risk stratification in ACHD patients.