AI Tool Matches Wire-Based FFR Accuracy in 36 Seconds

AI-Derived FFR Shows Strong Performance Against Invasive Standard

An artificial intelligence system that calculates fractional flow reserve values directly from coronary angiograms has demonstrated diagnostic accuracy comparable to invasive wire-based measurements, according to research published in JSCAI, the journal of the Society for Cardiovascular Angiography and Interventions.

The technology, AutocathFFR, was developed by Tel Aviv-based MedHub-AI and has received CE mark approval in Europe and regulatory clearance in Japan. It has not yet been approved by the U.S. Food and Drug Administration.

Researchers conducted a retrospective analysis of nearly 500 patients who underwent coronary evaluation between 2019 and 2023. The team applied AutocathFFR to existing angiogram images and compared the AI-generated values against wire-based FFR measurements taken during the original procedures.

Performance Metrics

The AI system produced a mean FFR measurement of 0.85, identical to the mean value obtained through wire-based techniques. Processing time averaged just 36.1 seconds per case.

For nearly three-quarters of patients (73.6%), the system automatically detected lesions without manual intervention. The remaining cases required semiautomated analysis with manual lesion marking.

Diagnostic performance metrics were strong across the board: sensitivity reached 90.2%, specificity hit 94.9%, and overall diagnostic accuracy stood at 93.8%. The area under the ROC curve was 0.93. Notably, the system maintained accuracy even for lesions with FFR values in the clinically ambiguous "gray zone" range.

Why It Matters

FFR assessment guides critical decisions about whether to perform revascularization procedures, but invasive wire-based measurement adds procedural time, cost, and requires adenosine administration. An accurate AI alternative that works with standard angiography could make physiologic assessment routine rather than selective, potentially improving treatment decisions while reducing procedural complexity. The 36-second processing time and automated operation could particularly benefit catheterization labs with high patient volumes.

Study Limitations and Next Steps

First author Eyal Ben-Assa, MD, of Assuta Ashdod University Hospital in Israel and the Massachusetts Institute of Technology, and colleagues acknowledged several limitations. The retrospective design meant researchers could not verify how carefully the original wire-based FFR measurements were performed. MedHub-AI, the company behind AutocathFFR, funded the analysis.

The research team emphasized that prospective studies are needed to determine whether the technology can effectively guide revascularization decisions and improve patient outcomes in real-world clinical settings.

The findings were first reported in JSCAI and detailed in Cardiovascular Business.