An Artificial Intelligence Algorithm for Detection of Severe Aortic Stenosis: A Clinical Cohort Study

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate identification of severe aortic stenosis (AS) patients at high risk of mortality is challenging with current imaging methods.
• Severe AS requires timely intervention to prevent adverse outcomes.

Purpose of the Study:

• To evaluate an artificial intelligence decision support algorithm (AI-DSA) for augmenting severe AS detection.
• To assess AI-DSA's performance in a large Medicare beneficiary cohort.

Main Methods:

• An AI-DSA was trained to recognize echocardiographic phenotypes associated with an aortic valve area (AVA) < 1 cm².
• The AI-DSA analyzed routine transthoracic echocardiogram (TTE) reports from 31,141 U.S. Medicare beneficiaries.
• The algorithm operated independently of clinical information and left ventricular outflow tract measurements.

Main Results:

• AI-DSA demonstrated excellent performance in detecting the severe AS phenotype (sensitivity 82.2%, specificity 98.1%, c-statistic 0.986).
• The algorithm identified an additional 3.3% of patients with moderate AS exhibiting a severe AS phenotype, who had low rates of aortic valve replacement.
• Five-year mortality was significantly higher in patients identified with severe AS or a similar phenotype compared to those without.

Conclusions:

• The AI-DSA reliably identifies the severe AS phenotype using echocardiographic reports without requiring left ventricular outflow tract measurements.
• This AI-DSA shows potential for improving the detection of severe AS and identifying at-risk individuals for better clinical management.