Electrocardiogram screening for aortic valve stenosis using artificial intelligence.

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Early detection of aortic stenosis (AS) is crucial for timely intervention.
• Improved outcomes are observed with aortic valve replacement in asymptomatic severe AS patients.
• Moderate AS carries a poor prognosis, highlighting the need for better screening.

Purpose of the Study:

• To develop an artificial intelligence-enabled electrocardiogram (AI-ECG) using a convolutional neural network.
• To identify patients with moderate to severe aortic stenosis (AS).

Main Methods:

• Utilized a large dataset of 258,607 adults from the Mayo Clinic database with echocardiography and ECG data.
• Trained, validated, and tested a convolutional neural network AI-ECG model on randomly selected subjects.
• Evaluated model performance using area under the curve (AUC), sensitivity, specificity, and accuracy.

Main Results:

• The AI-ECG model achieved an AUC of 0.85 in the test group, with 78% sensitivity and 74% specificity.
• Model performance improved with the addition of age and sex (AUC 0.87), further increasing to 0.90 without hypertension.
• False-positive AI-ECGs were associated with a 2.18-fold increased risk of developing moderate or severe AS within 15 years.

Conclusions:

• An AI-ECG can effectively identify patients with moderate or severe AS.
• AI-ECG demonstrates potential as a valuable screening tool for AS in community settings.
• Early identification via AI-ECG may facilitate timely management and improve patient prognosis.