Aortic Science Made Simple Evaluating AI to Detect Aortic Valve Disease

Aortic Science Made Simple, Issue 3

Title: Evaluating AI to detect Aortic Valve Disease

Aortic stenosis is the narrowing of the aortic valve– a cardiac outlet that allows for a one-way flow from the heart to the aorta and the rest of the body. As this opening becomes smaller, the heart must work harder to pump blood through it. In adults, this commonly develops as calcium accumulates on the valve leaflets, rendering them stiff and limiting their movement.

“The disease can develop gradually without obvious symptoms,” stated Dr.Ofir Koren, an interventional cardiologist and research scientist at the Smidt Heart Institute of Cedars-Sinai Medical Center. “Breathlessness, fatigue, or reduced exercise [common symptoms of aortic valve stenosis] capacity may be attributed to aging” continued Koren.

At the 2026 Society of Cardiovascular Computed Tomography (CT) meeting, Koren presented a novel AI-employed retrospective study identifying aortic calcifications. The study analyzed 440 chest CT scans from patients aged 50 and older who underwent imaging at Cedars-Sinai between January and March 2025 as part of their routine medical care. The scans themselves were taken without intravenous contrast dye and were not synchronized with the heartbeat. This made the study particularly relevant to routine imaging, which is not always optimized for examining the heart's structures.

The team used HealthAVC V1.0, a pretrained AI program developed by NanoxAI, designed to locate the aortic valve and measure the amount of calcium present. Then, “two experienced [CT] readers independently measured valve calcium, blinded to the AI results,” stated Koren. “[The CT readers’] measurements [were then compared] with the automated findings and reviewed whether valve calcification had been explicitly documented in the original radiology reports.”

Among the 440 scans, the CT readers confirmed that 172 showed aortic valve calcium above the study's predefined threshold. However, only 27 out of 172 scans mentioned aortic valve calcium in the original radiology report. In other words, approximately 84% of the confirmed positive findings were not specifically documented in the reports.

However, HealthAVC V1.0 identified 169 of the 172 confirmed positive cases, corresponding to a sensitivity of 98.3%. In medical research, “sensitivity” describes how well a test correctly identifies cases in which the finding being investigated is actually present. In this study, the high sensitivity suggests that the AI software detected calcium in almost all of the scans confirmed –by the CT readers– to have calcium accumulation.

“The main finding was that valve calcium was common in this [study’s] population, frequently absent from the written reports, and detectable with AI in most scans,” said Koren. While HealthAVC V1.0 does not diagnose aortic stenosis itself, AI detention of valve calcification could allow for earlier assessment and followup before significant disease progression. However, “It does not mean that everyone with valve calcium needs an intervention or will develop severe stenosis,” Koren emphasized.”The amount of calcium, valve function, symptoms, and overall clinical context all matter.”

For Koren and his team, the next step is to evaluate whether AI detection can improve care in practice. “Longer-term studies should assess whether that pathway leads to more timely diagnosis, better follow-up, and improved patient outcomes.”

Thank you for reading. ~Aditi Dorokdar