AI-supported scans measuring heart fat could better predict cardiovascular risk
A new study suggests that AI can measure heart fat from routine coronary artery calcium (CAC) scans without requiring additional tests.
Higher levels of this heart fat were independently linked to a greater risk of developing cardiovascular disease over long-term follow-up.
Adding the AI-derived heart fat measurement to existing risk models could significantly improve the accuracy of cardiovascular risk prediction.
The study indicates this improvement may be especially useful for people at low or intermediate risk, helping better identify those who may benefit from earlier preventive care.
Cardiovascular diseases are the leading cause of death globally, and more than 60% of United States adults have at least one risk factor for these conditions.
Early diagnosis is crucial for managing the condition, preventing irreversible heart damage, and reducing hospitalization. However, early diagnosis can be challenging, as many heart diseases often develop silently without noticeable symptoms until advanced stages.
Coronary artery calcium (CAC) scans are a routine imaging test that measures calcium in the coronary arteries and can detect early signs of heart disease.
It is a quick and noninvasive procedure that can help predict an individual’s cardiovascular disease risk. As technology continues to improve, artificial intelligence (AI) techniques are showing great potential for improving the accuracy, efficiency, and timing of cardiovascular disease diagnosis.
Now, a new study suggests that using AI to measure fat around the heart, known as pericardial fat, using CAC scans could significantly improve the ability to predict a person’s risk of developing cardiovascular disease.
Prior research highlights a strong association between pericardial fat volume and cardiovascular disease.
The findings, presented at the American College of Cardiology Scientific Session 2026 and published in the American Journal of Preventive Cardiology, highlight how AI can extract additional clinically useful information from routine imaging tests.
What did the study find?
The study followed nearly 12,000 adults who underwent CAC scans for approximately 16 years to track the development of cardiovascular disease. The researchers used AI to analyse participants’ scans and measure the fat surrounding the heart.
They compared the predictive value of this measurement with and in combination with two standard risk assessment approaches.
This included the American Heart Association (AHA) PREVENT equation, which incorporates factors such as age, sex, blood pressure, cholesterol, and diabetes, and the coronary artery calcium score, which measures calcified plaque in coronary arteries.
“Current risk prediction tools categorize a meaningful proportion of patients as borderline or intermediate risk; our study shows that this automated biomarker can identify higher risk individuals within those categories that may benefit from earlier or more aggressive preventive treatments and intervention,” noted Lopez-Jimenez.
“And importantly, this will not require any additional imaging beyond what is already being done for the patients,” he added.
Pericardial fat as a predictor of cardiovascular outcomes
Notably, the results suggest that pericardial fat volume can be used independently to predict cardiovascular events.
Importantly, heart fat volume remained a predictor of risk even after accounting for traditional factors such as age, blood pressure, cholesterol, and diabetes.
This measurement also improved prediction accuracy when combined with the existing risk models. The benefit was particularly notable in those considered low or intermediate risk.
“Pericardial fat’s contribution to predicting cardiovascular outcomes was previously shown in several other studies,” said Zahra Esmaeili, MD, first author and researcher in the Department of Cardiovascular Medicine at Mayo Clinic.
“However, what was notable to us was that this biomarker can add incremental values on top of both traditional risk factors, and coronary calcium scoring, and beyond current risk assessment tools,” Esmaeili noted.
“Specifically, higher pericardial fat volume provided increased value in borderline and intermediate risk patients and showed a 24% higher risk among individuals with low coronary calcium,” she added.
Why does heart fat matter?
Pericardial fat has long been recognized as a marker of cardiovascular risk. This type of fat is thought to play an active role in heart disease through inflammatory and metabolic processes that may affect nearby coronary arteries.
For example, research highlights that pericardial fat can significantly increase the risk of heart failure and is also linked to higher risks of coronary artery disease and myocardial dysfunction.
However, measuring pericardial fat is not routine in clinical practice, as measuring it manually has been time consuming and impractical.
Therefore, AI may enable this measurement by offering automated, rapid, and consistent analysis of imaging data.
“Pericardial fat is visible on routine coronary artery calcium scans, but measuring it manually for each patient is time-consuming and prone to variability depending on who is doing the measurement,” Lopez-Jimenez explained.
“Our AI model was trained on a set of manually annotated images, and it learned to automatically identify and segment this fat depot with high accuracy; and then it provides the volume of the segmented parts of the images,” he added.
Improving existing risk tools
Clinicians currently estimate cardiovascular risk using established models, such as the PREVENT equation, alongside CAC scores.
However, while these approaches are better calibrated than previous methods, they lower risk estimates and reclassify many borderline patients. As such, they may be less clinically precise for those in intermediate-risk categories, potentially leading to shifts in treatment decisions.
The researchers suggest a significant improvement in long-term risk prediction when combining the AI-derived heart fat measurements with the traditional tools. This may help clinicians to make more informed decision about when to start preventive treatments.
“The groups most likely to benefit are those in the borderline and intermediate PREVENT risk categories, where the decision to initiate or intensify preventive therapy is more uncertain,” Esmaeili told MNT.
“Similarly, patients with zero or low coronary calcium scores may carry residual cardiometabolic risk that pericardial fat volume can help uncover,” she said. “Additionally, our analyses showed that higher pericardial fat is prognostic of cardiovascular events in patients with normal body mass index, this highlights the importance of visceral adiposity in normal weight individuals.”
While the findings add to a growing body of research showing how AI could improve cardiovascular risk assessment and detection, further studies are still necessary to determine how best to integrate AI-derived pericardial measurements into routine clinical practice.
References
AI-supported scans measuring heart fat could better predict cardiovascular risk, medical news today, www.medicalnewstoday.com/articles/ai-scans-measuring-heart-fat-better-predict-cardiovascular-risk