AI Tool Detects Hidden Heart Risk in Routine Chest CT Scans

A new artificial intelligence tool is helping doctors detect early warning signs of heart disease in routine chest CT scans, potentially identifying patients at risk years before symptoms appear.

Developed by researchers at Mass General Brigham in collaboration with the U.S. Department of Veterans Affairs, the tool—called AI‑CAC—uses deep learning to spot coronary artery calcium (CAC), a mineral marker of past plaque damage that has been linked to higher risk of heart attacks and death. A calcium score over 400, for instance, can triple a person’s 10-year risk of heart attack or death.

The technology enables faster, more consistent readings from scans that are often taken for unrelated reasons, such as evaluating lung or spine issues.

Missed Clues in Everyday Scans

Millions of chest CT scans are performed each year, but because they’re not typically intended to evaluate the heart, early signs of heart disease often go unnoticed. These routine, "nongated" scans do capture the heart, but radiologists usually focus on other areas and don’t report coronary calcium unless specifically asked.

That gap in care means many patients don’t learn they’re at risk until after symptoms appear—when preventive measures like statins or lifestyle changes have less time to make a difference.

How AI‑CAC Works

The team trained AI‑CAC on chest scans from 98 Veterans Affairs hospitals, covering a broad range of body types and CT scanner models. The deep learning system taught itself to identify patterns of calcium buildup, using labeled scans to learn how to score them automatically.

In a test involving over 8,000 scans, AI‑CAC detected the presence of any coronary calcium with 89% accuracy. It also correctly identified whether a patient's calcium score was above or below 100 in 87% of cases—performance on par with human experts, but completed in seconds rather than minutes.

“Millions of chest CT scans are taken each year, often in healthy people,” said study lead Dr. Hugo Aerts. “Teaching the algorithm to comb existing archives can enable physicians to engage with patients earlier, before their heart disease advances to a cardiac event.”

Calcium Scores and Risk

Calcium in the coronary arteries is considered a warning sign for heart disease, even when patients have no symptoms. A score over 400, for example, is associated with a threefold increase in the 10-year risk of heart attack or death.

In the study, patients with AI‑CAC scores above 400 had a 3.49 times higher risk of dying over the next decade compared to those with zero calcium. When cardiologists reviewed a random sample of these high-score scans, they agreed in 99% of cases that the patients would benefit from starting cholesterol-lowering medication.

A Scalable, Inclusive Approach

What makes this effort notable is its scale and diversity. Many earlier AI models were trained on small, homogenous datasets using “gated” heart-specific scans from a single scanner type. By contrast, AI‑CAC was developed using millions of real-world, nongated scans from a wide mix of VA hospitals.

Still, the VA population is predominantly older and male. Co-author Dr. Raffi Hagopian acknowledged this limitation, but called it a strong starting point. “VA imaging archives already hold millions of nongated chest CT scans,” he said, calling it a valuable base for expanding the model into broader healthcare settings.

Academic medical centers are now planning head-to-head validation studies later this year to test AI‑CAC in more diverse populations, using a range of scanners and imaging protocols.

From Detection to Prevention

The technology aligns with existing medical guidelines. The American College of Cardiology and American Heart Association already recommend starting statins in select patients with CAC scores above 100. But that usually requires a specialized, gated scan—which isn’t always ordered or covered by insurance.

AI‑CAC circumvents that hurdle by using existing scans, no extra imaging required. If further trials confirm its accuracy, it could be deployed through a simple software update to hospital radiology systems, helping flag at-risk patients before they leave the imaging suite.

“Using AI for tasks like CAC detection can help shift medicine from a reactive approach to the proactive prevention of disease,” said Hagopian.

Challenges and Limitations

Automating coronary calcium detection offers clear benefits—reducing the workload for radiologists and identifying high-risk patients before they even leave the imaging suite. But it also raises important considerations for implementation.

Clinical protocols will be essential. False positives could lead to unnecessary anxiety or follow-up testing, so hospitals will need guidelines to determine when and how to act on AI-generated alerts.

Data privacy and security are also key. The model was developed within the secure environment of VA hospitals, which operate behind strict firewalls. Expanding to private healthcare systems will require equally robust protections to safeguard patient information.

There are also legal and operational questions to resolve, such as who holds liability if the algorithm produces an inaccurate result, and how to reimburse the additional patient counseling that may follow each AI-generated risk flag.

What's Next

Looking ahead, the developers plan to test AI‑CAC in community hospitals and monitor whether early statin prescriptions lead to better outcomes. Future efforts will also aim to integrate calcium scores directly into electronic health records, alongside existing metrics like blood pressure and cholesterol—bringing AI‑powered heart risk assessments closer to routine clinical care.

What This Means

AI‑CAC could mark a shift in how heart disease risk is identified—by turning ordinary chest scans into a powerful early warning system. This approach could reduce missed opportunities for preventive care, especially among patients who might never receive a dedicated heart scan.

If validated across broader settings, AI‑CAC has the potential to reshape how cardiovascular risk is detected and managed—without requiring new scans or added radiation. By turning routine imaging into a proactive screening tool, it opens the door to earlier intervention, more personalized care, lower healthcare costs, and a system that identifies risk before symptoms appear.

For now, the study shows what’s possible when overlooked clinical data meets practical AI—and puts a proven marker of heart risk back into the hands of patients and their doctors.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.