Reviewer: : Dr. Onur Akhan

Name of the Study: Artificial intelligence identifies severe aortic stenosis from routine echocardiograms (AI-ENHANCED AS study)

Published Congress: ESC 2022

Link : https://www.escardio.org/The-ESC/Press-Office/Press-releases/Artificial-intelligence-identifies-severe-aortic-stenosis-from-routine-echocardiograms

Background:

In Europe and North America, the most common primary valve lesion that requires surgery or transcatheter intervention is aortic stenosis, and its prevalence increases with aging. The current guidelines advise early intervention for symptomatic severe aortic stenosis patients. Echocardiography is the most valuable diagnostic tool. However, mortality risk increases beyond the current diagnostic definitions, and consequently, more patients should be evaluated for this purpose.

Objective:

The study aims to demonstrate to what extent the echocardiography database, using artificial intelligence assistance, can be routinely used in clinical practice to identify moderate-to-severe and severe aortic stenosis phenotypes with an increased risk of five-year mortality.

Methods:

The proprietary AI-Decision Support Algorithm (AI-DSA) used has been associated with mortality information by examining more than 1,000,000 the National Echo Database of Australia (NEDA) echocardiogram data in more than 630,000 patients. The algorithm was also trained using randomly selected 70% NEDA data to ensure that all severe aortic stenosis was detected as defined by the guidelines. The remaining 30% of NEDA data compared the five-year mortality rates between moderate-to-severe and severe aortic stenosis phenotypes and without severe aortic stenosis phenotypes.

Results:

The algorithm identified 1.4% as a moderate-to-severe phenotype and 2.5% as a severe phenotype; in those with a severe phenotype, 77.2% met the guideline criteria for severe aortic stenosis with a five-year mortality rate of 69.1%. A five-year mortality rate of 64.4% was observed in the remaining part that did not meet the guideline criteria. (The five-year mortality rate for the moderate-to-severe phenotype was 56.2%, for the severe phenotype 67.9%, and others, 22.9% due to guidelines.)

Conclusion:

In light of the study's data, algorithm aid can be considered in diagnosing high-risk patients that traditional diagnostic methods cannot detect.

Interpretations:

When examining the increasing prevalence of aortic stenosis and its impact on mortality, new methods should be considered to reconsider the diagnosis of patients and identify those at risk. AI-DSA algorithm is one of them, but more research is needed.