المتخصصون في المجالات الطبية

Adding AI-derived pericardial adipose tissue measurement helps predict cardiovascular disease risk beyond coronary calcium

May 19, 2026

Mayo Clinic Cardiovascular Medicine researchers discovered a new independent predictor of cardiovascular disease (CVD) risk in patients. The artificial intelligence (AI)-based measurement of pericardial adipose tissue (PAT) volume from an electrocardiogram (ECG)-gated coronary artery calcium (CAC) CT scan helps forecast CVD events.

This deep learning (DL)-based quantification of PAT volume can be incorporated into risk assessment. And the measurement comes from imaging that many patients already receive.

The large U.S. population-based cohort study was published in the American Journal of Preventive Cardiology. Findings were presented at the 2026 American Academy of Cardiology Scientific Session.

Building better DL-derived tools

The researchers investigated the DL-based measurement of PAT in combination with the two standard CVD risk assessment approaches:

  • Coronary artery calcium (CAC) score measures the calcified plaque in the coronary arteries from an ECG-gated CT scan of the heart.
  • American Heart Association PREVENT equation that predicts risk of CVD events using age, blood pressure, cholesterol, diabetes, sex and other factors.

"Current evidence shows that about one-third of cardiovascular events happen in people with zero CAC score. This means that in using the tool we rely on most we tend to miss a subset of patients who will have events in the future," says Zahra Esmaeili, M.D., a Cardiovascular Medicine research fellow at Mayo Clinic in Rochester, Minnesota, and first author of the study. Dr. Esmaeili works in Mayo's Artificial Intelligence (AI) in Cardiovascular Medicine specialty group. The AI in Cardiovascular Medicine team has been developing AI cardiovascular risk prediction models for years. Several models have been approved for clinical use.

Using pericardial biomarkers

The investigators analyzed data of 11,897 adults (65% male) from Olmsted County, Minnesota, without known history of heart failure and atherosclerotic cardiovascular disease events who underwent noncontrast, ECG-gated CAC CT scans. Over a median of 16.4 years, 9.6% experienced CVD events.

"Pericardial adipose tissue is the fat surrounding the coronary arteries in direct paracrine contact with cardiac structures. Because of this proximity and that there's no fascial barrier separating it from the myocardium, PAT creates a local inflammatory environment around the heart. It likely carries biological information about future events that traditional risk factors simply don't capture," Dr. Esmaeili says.

An emerging imaging biomarker of CVD risk, PAT has already been recognized as a risk factor for cardiovascular events based on both manual and automated CT measurements.

"The field of AI-derived imaging biomarkers to fill gaps in current risk assessment tools is growing fast. We went further than asking whether there's an association between PAT and cardiovascular events. We looked at where this connection can help the most in clinical practice," says Francisco Lopez-Jimenez, M.D., M.S., a preventive cardiologist at Mayo Clinic in Rochester, Minnesota, co-director of Artificial Intelligence (AI) in Cardiovascular Medicine and senior author of the study. "What made this study clinically useful was accessing a large longitudinal population-based cohort at Mayo Clinic using Rochester Epidemiology Project data linkage. This scale of data access brought meaning to using an AI-derived result in a real-world setting to investigate the incremental value of this marker over two established risk tools."

Recognizing residual risk

Looking at PAT when combined with the standard risk assessment approaches improved overall accuracy of long-term risk prediction, especially in patients in lower risk categories.

"When we added PAT volume and density to a model that already included PREVENT risk factors and CAC score, the C-statistic went from 0.72 to 0.76. That's a significant improvement," Dr. Esmaeili says.

Patients in lower risk categories benefit most from this tool, including:

  • People in borderline and intermediate PREVENT risk categories.
  • Patients with lower CAC scores below 100 Agatston units.

Identifying patients traditionally considered low risk who are at risk of future adverse events addresses a void in prior studies.

Study highlights:

  • PAT volume is associated with future events, not density. Volume in the highest tertile was associated with 30% higher CVD risk after accounting for PREVENT risk factors and CAC score.
  • Higher PAT volume leads to higher risk. The dose-response relationship was linear, and restricted cubic spline analysis showed no J- or U-shaped curve.
  • PAT volume could stratify patients for future events even in individuals with a CAC score of zero or below 100 Agatston units.

"This tells us something vital. We typically consider patients with a zero CAC score low risk and reassure them. The risk of future cardiovascular events is heterogenous. The 7.3% adverse event rate we observed in these patients over 16.4 years shows this in our cohort. PAT helps us to identify who is carrying residual risk that the calcium score alone is missing," Dr. Esmaeili says.

Next steps

Whether the results persist in different population subgroups requires more evidence. The cohort for this study was 97% white.

"We are particularly interested in exploring whether our findings persist in other populations who may appear low risk by current measures but carry a hidden elevated risk that PAT can help unmask," Dr. Esmaeili says. "And specifically, patients within normal weight when their visceral and pericardial fat burden may tell a different story. It's another population where this tool could have clinical impact."

Following prospective validation, the next step is testing whether reducing PAT volume can change outcomes through:

  • Statins.
  • GLP-1 agonists.
  • SGLT2 inhibitors.
  • Lifestyle intervention.

"The most direct influence would be in how we handle patients in that gray zone. Individuals with borderline or intermediate PREVENT risk scores and CAC scores below 100, where the decision to start or intensify preventive therapy is uncertain in absence of other comorbidities," Dr. Lopez-Jimenez says. "Measuring PAT volume automatically from a noncontrast CT scan could offer information for managing patients more precisely."

For more information

Esmaeili Z, et al. Deep learning-derived pericardial adipose tissue by electrocardiogram-gated cardiac computed tomography predicts cardiovascular events beyond coronary calcium score. American Journal of Preventive Cardiology. In press.

Refer a patient to Mayo Clinic.