An artificial intelligence-driven angiography-based FFR (MPFFR) demonstrated comparable diagnostic accuracy with quantitative flow ratio (QFR) in identifying functionally significant stenosis, and similar prognostic ability with QFR in terms of target vessel failure at 2 years.
MPFFR required a mean analysis time of 12.5 seconds with manual correction needed in only 5.3% of vessels.
MPFFR showed a statistically stronger correlation with FFR than QFR, though both performed similarly.
MPFFR and QFR had statistically similar area under the curve (AUC) for predicting functionally significant stenosis defined by FFR ≤0.80.
Patients with MPFFR ≤0.80 had significantly higher risk of target vessel failure at 2 years compared to those with MPFFR >0.80.
MPFFR and QFR had comparable prognostic ability for predicting target vessel failure at 2 years.
The study prospectively enrolled 599 vessels from 452 patients across 5 university hospitals in Korea who underwent clinically indicated FFR measurement.
This research suggests that an artificial intelligence (AI)-powered tool called MPFFR can rapidly and accurately assess whether narrowed coronary arteries are causing significant blood flow restriction — a key question in deciding whether patients need interventions like stenting. Unlike traditional methods that require cardiologists to manually trace artery images, MPFFR automates this process in about 12.5 seconds on average, requiring manual correction in only about 5% of cases. The AI tool's accuracy was compared against both an established wire-based pressure measurement (FFR, the gold standard) and another established software tool (QFR), with results showing the AI approach performed comparably to the existing technology. In a follow-up of roughly 2 years across 599 coronary vessels in 452 patients at five Korean hospitals, patients whose arteries were flagged as having restricted flow by MPFFR were nearly six times more likely to experience serious cardiac events (including cardiac death, heart attack, or need for repeat procedures) than those whose arteries were not flagged. This suggests the tool not only diagnoses the immediate problem accurately but also helps identify patients at higher long-term risk. This research matters because assessing coronary artery blockages currently requires specialized equipment and time-consuming manual steps. An AI-driven approach that can perform this analysis in seconds — with high accuracy and without needing a separate pressure wire inserted into the heart — could make this type of functional assessment faster and more accessible, potentially improving decision-making for patients with coronary artery disease.
Lee S, Gim D, Hwang D, Kim H, Cho H, Kim S, et al.. (2026). Artificial Intelligence-Driven Angiographic Quantification of Fractional Flow Reserve: Proof-of-Concept Validation Study.. Journal of the American Heart Association. https://doi.org/10.1161/JAHA.125.049331