What if a routine coronary artery calcium scan could see aortic stenosis risk before symptoms?

ESC 2026 · 10 Scientific Presentations · Oral & Poster · 7/10

One CT. 25 automated phenotypes. Nearly two decades of follow-up.

Aortic stenosis is often recognized only after obstruction or myocardial damage has already developed. But could information already captured in a routine noncontrast cardiac CT help identify risk much earlier?'

In a new HeartLung.AI scientific poster at ESC Congress 2026, researchers explored whether AI-CVD-derived imaging phenotypes could predict incident aortic stenosis long before clinical recognition.

The study included 5,520 MESA participants, with 157 incident cases of aortic stenosis over approximately two decades. AI-CVD automatically evaluated 25 imaging measures—including valvular and vascular calcification, cardiac chamber structure, adipose tissue, muscle, and bzone.

The results

The AI-CVD-AS model achieved:

0.973 AUROC for 10-year discriminationvs. 0.809 for the clinical model

Among the strongest imaging predictors were:

• Aortic valve calcification (AVC)

• Mitral valve calcification (MVC)

• Thoracic aortic calcification (TAC)

• LV mass index

• AI-derived coronary artery calcium

A simple five-variable model using age, sex, AVC, MVC, and TAC substantially outperformed the clinical model.

These findings suggest automated phenotyping of routine noncontrast cardiac CT may offer an opportunity for earlier, opportunistic assessment of aortic stenosis risk - without requiring an extensive clinical risk-factor panel.

External validation and recalibration are required before clinical implementation.

📍 To be presented at ESC Congress 2026 · Munich

By Zahra Heidari Meybodi M.D.