A new, AI-based video biomarker can highlight the risk of aortic stenosis or its future progression on cardiovascular scans, a study indicates. This study opens up the possibility to opportunistically evaluate patients on handheld devices.
The Digital Aortic Stenosis Severity index (DASSi) could help personalize the trajectory of individuals whose aortic valve disease appears superficially similar on echocardiograms, according to the research in JAMA Cardiology.
The imaging-based algorithm is trained to detect general features associated with severe aortic stenosis and can identify patients who do not meet traditional criteria yet exhibit fast rates of progression.
It was able to identify an elevated risk of aortic stenosis, its development and the need for future aortic valve replacement among nearly 60,000 patients assessed.
“By practical deployment in any setting where video-based cardiac imaging is obtained, DASSi may enable more precise risk stratification for the most common valvular disorder without any changes in the image acquisition protocols,” the researchers report.
Transcatheter and surgical aortic valve replacement alters the prognosis of aortic stenosis, and their increasing availability has led to rising interest in how to identify the risk of rapid progression.
However, this has been hampered by the high prevalence of milder forms of aortic valve disease, such as aortic sclerosis, with personalized follow up further impacted by variability in the progression rates among patients with similar Doppler-assessed severity.
Existing biomarkers do not inform for personalized screening and follow-up, and monitoring is currently dependent on referrals for comprehensive testing by Doppler echocardiography.
Noting the need to better define the personalized trajectory of aortic stenosis, Evangelos Oikonomou, PhD, from Yale School of Medicine, and co-workers assessed the value of a deep learning strategy they developed.
This involves recognizing the computational representation of severe aortic stenosis on simple parasternal long-axis videos without Doppler characterization, which is a standard echocardiographic view easily captured on handheld device.
The team studied a total of 12,599 patients without severe aortic stenosis, all at least 54 years of age, who were undergoing echocardiography at either Yale New Haven Health System or Cedars-Sinai Medical Center and were followed up for a median of 4.1 and 3.4 years, respectively.
Higher baseline DASSi was associated with faster progression in the annualized change in peak aortic valve velocity at both centers, at a corresponding 0.033 and 0.082 m/s per year for every 0.1 DASSi increment.
DASSi values of at least 0.2 were associated with a four- to five-fold higher risk of aortic valve replacement compared with values of less than this, independently of potential confounding factors.
The findings were reproduced among 45,474 individuals using cardiac magnetic resonance videos in the U.K. Biobank.
Here, having a DASSi of at least 0.2 versus less than 0.02 was associated with a hazard ratio of 11.38 for aortic valve replacement.
The researchers maintain: “These findings support the use of DASSi, a cross-modal artificial intelligence biomarker for both opportunistic [aortic stenosis] screening on echocardiography as well as deeper phenotyping of [aortic stenosis] on standard noninvasive modalities and potentially handheld devices.”
