New Artificial Intelligence Tool Successfully Detects Severe Aortic Stenosis Using Routine Echocardiogram
In a new study published in JACC: Advances, researchers at Beth Israel Deaconess Medical Center (BIDMC) successfully validated an artificial intelligence tool that detects severe aortic stenosis (AS) with over 98 percent accuracy, a common and serious valve disease affecting about five percent of adults older than 65 in the U.S.
Jordan B. Strom, MD, MSc, Head of Cardiovascular Imaging Research at the Richard A. and Susan F. Smith Center for Outcomes Research at BIDMC, and colleagues tested the artificial intelligence decision support algorithm (AI-DSA) using routine echocardiographic reports from more than 31,000 U.S. Medicare beneficiaries at BIDMC. The AI-DSA identified 98 percent of patients who met clinical guidelines for severe AS and identified more than a thousand patients who likely had severe AS, but did not meet clinical guidelines. These patients had the same risk of death as those with severe AS, but only 6.6 percent received an aortic valve replacement (AVR), a life-extending treatment, versus 20.2 percent of people diagnosed with severe AS.
The results of this study suggest a potential use for this algorithm alongside standard echocardiographic reporting to improve identification of individuals with severe AS symptoms and subsequent referral for lifesaving AVR. Moving forward, the AI-DSA tool should undergo clinical testing to ensure that its use improves existing clinical practice and overall detection of individuals with severe AS.
Aortic stenosis (AS) is a serious heart valve disease that narrows the aortic valve, reducing blood flow from the heart to the aorta. It is the most common form of valvular heart disease encountered by clinicians. Severe AS is almost universally fatal without undergoing aortic valve replacement (AVR). Despite guideline-based indications to refer individuals with symptomatic, severe AS to AVR and an increased mortality risk of 2 percent for each week that referral is delayed, nearly one in three of these individuals are not referred.
Originally trained and tested using over one million echocardiograms from the National Echocardiographic Database of Australia, the AI-DSA studied uses echocardiographic reports to predict an individual's probability of having severe AS without relying on commonly used, error-prone measurements that can under-detect severe AS.
Read the full paper in JACC Advances.
BILH Study Authors: Jordan Strom, MD, MSc,
COI: This work was supported by EchoIQ. Jordan B. Strom, MD, MSc, serves on the board of scientific advisors for EchoIQ.
Citation: Strom, J, Playford, D, Stewart, S. et al. An Artificial Intelligence Algorithm for Detection of Severe Aortic Stenosis: A Clinical Cohort Study. JACC Adv. 2024 Sep, 3 (9_Part_2). https://doi.org/10.1016/j.jacadv.2024.101176
