What New CCTA and Deep Learning Research Reveal About Small Plaque Volume and Coronary Atherosclerosis

The smaller the plaque, the bigger the message.For decades, cardiovascular prevention has been built around risk factors, calcium, and stenosis. This new SCAPIS analysis suggests that we may be missing the most obvious variable: How much coronary atherosclerosis is actually there?¹

In 23,314 asymptomatic adults without known atherosclerotic cardiovascular disease (ASCVD), deep learning automatically quantified total plaque volume (TPV) and non-calcified plaque volume on coronary computed tomography angiography (CCTA). Over a median of 7.8 years, 287 coronary events occurred.¹

The relationship was strikingly dose-dependent.

In comparison with no detectable plaque, progressively higher TPV was associated with progressively higher risk, reaching an adjusted hazard ratio (HR) of 6.36 in the highest plaque-volume category even after accounting for Systematic Coronary Risk Evaluation 2 (SCORE2) and disease extent. Adding TPV to SCORE2 + segment involvement score (SIS) improved discrimination (C-statistic 0.783 → 0.798) and risk reclassification.¹

What is the Critical Point with the New SCAPIS Analysis?

The most interesting result is not the AI.
It is what the AI was able to see.

Among individuals with a coronary artery calcium (CAC) score of 0, even relatively small TPV (> 0–32.1 mm³) was associated with a 2.63-fold higher risk, rising to 5.26-fold when TPV exceeded 32.1 mm³.¹

Even more provocatively, the algorithm detected plaque in 73 percent of CAC=0 individuals and 72 percent of those considered plaque-free by manual CCTA assessment, with risk increasing progressively with plaque volume.¹ Adding ≥ 50 percent stenosis provided essentially no additional predictive information.

Does This Study Represent Another Step Away from Stenosis-Centered Cardiology?

The hierarchy increasingly looks like this:

• Risk factors predict probability.
• CAC detects established calcified disease. 
• CCTA detects plaque. 
• Quantitative CCTA measures the disease burden.

AI matters because it makes this quantification scalable. However, the real biomarker is not the algorithm. It is the plaque.

Where Photon-Counting CT Fits into the Equation

Now imagine applying the same approach to photon-counting CT.

Higher spatial resolution, reduced blooming and improved characterization of very small calcified and noncalcified components could provide a substantially richer substrate for automated plaque quantification.

This study also contains an important clue. Even plaque volumes as small as > 0–11 mm³ tended toward approximately twice the event risk.¹

When increasingly tiny amounts of disease carry prognostic information, spatial resolution stops being merely an image-quality metric. It becomes part of risk assessment.

Perhaps the future of prevention is not: calculate risk → treat risk factors. Perhaps the path forward should be: detect plaque → quantify plaque → characterize plaque → treat the disease.

Dr. Cademartiri is the director of advanced cardiovascular imaging and photon-counting CT at the Scientific Institute for Research, Hospitalization, and Healthcare Synlab Diagnostic Network in Naples, Italy. He is also a consultant in advanced cardiovascular imaging at CDI/Centro Diagnostico Italiano in Milan, Italy.

(Editor’s note: This blog is adapted with permission from Dr. Cademartiri’s original LinkedIn post at: https://www.linkedin.com/feed/update/urn:li:activity:7501166309393473536/

Reference

1. Malmqvist J, Wang C, Bergstrom G, et al. Incremental predictive value of deep learning-quantified coronary atherosclerotic volume: a SCAPIS cohort analysis. JACC Cardiovasc Imaging. 2026 Aug 21:S1936-878X(26)00438-9.