A noninvasive artificial intelligence (AI) tool may be able to predict potential cardiovascular events in patients with suspected stable coronary artery disease.

Known as CT-derived fractional flow reserve (FFR-CT), the tool can help analyze images from coronary CT angiographies that are used to determine coronary artery disease, according to a large observational cohort study presented at the European Association of Cardiovascular Imaging (EACVI) 2025  conference. The tool assesses coronary blood flow, an essential diagnostic component of angina.

The study, known as FISH&CHIPS, is the first to provide conclusive evidence of FFR-CT’s prognostic power, in addition to its diagnostic abilities, said study presenter Jack Bell, MD, of the Liverpool Heart and Chest Hospital in Liverpool, England. The largest real-world FFR-CT cohort to date, it included nearly 8000 patients undergoing FFR-CT analysis, he said. The patients were from 25 hospitals.

FISH&CHIPS was based on a substantial number of clinical events, enabling analysis of individual outcomes such as myocardial infarction (MI), cardiovascular death, and all-cause death rather than composite endpoints. The study had a follow-up period of 3 years, longer than most prior FFR-CT studies, Bell said.

No Further Testing Required

“FFR-CT is already well established for the diagnosis of flow-limiting coronary artery disease that causes angina,” Bell said. “The advantage of this technique is that it doesn’t require any further testing, which is beneficial for patients and reduces the burden on the healthcare system. This prognostic value is as equally important as the diagnostic value in guiding treatment decisions.”

The study included 7836 patients (median age, 63 years; 37.4% women). After follow-up, 2.4% of patients experienced MI, 20.1% revascularizations, and 0.9% cardiovascular death and 3.3% died from any cause.

When FFR-CT was measured close to the stenosis site, researchers observed that lower values were linked to higher rates of cardiovascular events. Bell explained that if there is severely reduced stenosis-specific FFR-CT, there is a 5.7-fold greater risk for MI and a 3.4-fold risk for cardiovascular death within 3 years compared with a normal stenosis-specific FFR-CT. He added that this was independent of traditional cardiovascular risk factors such as age, sex, hypertension, diabetes, and dyslipidemia.

AI-based FFR-CT gives clinicians a clearer sense of which coronary lesions are likely to impair blood flow, not just whether the vessel appears narrow visually, said Rohan Khera, MD, MS, an associate professor at Yale University and director of the Cardiovascular Data Science Lab, both in New Haven, Connecticut.

Strengthened Decision-Making

“This helps move care from a purely anatomical assessment to one that reflects physiologic significance,” said Khera, who was not involved in the study. “For patients, it strengthens decision-making by helping identify who is at higher risk and who may be safely managed noninvasively.”

When FFR-CT indicates preserved flow, Khera explained, clinicians can often defer invasive angiography with confidence. “When it identifies physiologic compromise, evaluation can be directed more efficiently,” he said. 

“Adding prognostic context to the tool’s ability to diagnose disease can further provide reassurance when decisions are being made based exclusively on a noninvasive test.”

Khera said the gradations in FFR-CT severity correlate with different levels of risk, “which supports individualizing therapy intensity, follow-up frequency, and decisions around revascularization. However, there will need to be future evaluations of FFR-CT-guided treatment paradigms.”

In routine practice, clinicians adjust management once FFR-CT identifies impaired blood flow, Khera pointed out. 

“Since this is an observational study and reflects care pathways influenced by the test itself, the reported event rates likely represent a combination of underlying disease risk and the modifying effects of subsequent treatment,” Khera said. “Therefore, based on what is reported, analyses describe outcomes l inked with FFR-CT categories rather than the untreated natural history of the disease.”  

Bell concluded that using coronary CT angiography enables cardiologists to observe the disease process that leads to heart attacks, which differs for each patient. “With a growing number of options to treat coronary artery disease, it will be important that patients receive personalized risk assessment and targeted treatment to reduce this risk,” he said.

Bell reported having no financial disclosures. Khera reported being an associate editor of JAMA. He reported receiving support from the National Institutes of Health, the Doris Duke Charitable Foundation, and the Blavatnik Family Foundation. Khera also reported receiving research support, through Yale University, from Bristol Myers Squibb, Novo Nordisk, and BridgeBio. He reported being a co-inventor of several US Pending Patent Applications and being a co-founder of Ensight-AI, Inc., and Evidence2Health, two health platforms that aim to improve cardiovascular diagnosis and evidence-based cardiovascular care.

Martta Kelly is a medical journalist who lives in the New York metropolitan area.