New AI-powered FFR tech shows potential
A new advanced artificial intelligence (AI) tool out of Israel can deliver fractional flow reserve (FFR) values from routine coronary angiograms that are comparable to those captured by invasive wire-based techniques. That’s according to new data published in JSCAI, the flagship journal of the Society for Cardiovascular Angiography and Interventions.[1]
The technology at the heart of this analysis, AutocathFFR, was developed by Tel Aviv, Israel-based MedHub-AI. The automated tool has not yet received U.S. Food and Drug Administration approval. It has received CE mark approval, however, in addition to being approved in Japan.
This retrospective study focused on nearly 500 patients originally evaluated from 2019 to 2023. Researchers used AutocathFFR to review coronary angiograms taken when these patients were first evaluated; the new AutocathFFR values were then compared to wire-based FFR values taken at the time.
Image courtesy of JSCAI.
Overall, the mean AI-FFR measurement was 0.85—this is exactly the same as the mean wire-based FFR. The mean processing time for AI-FFR, meanwhile, was just 36.1 seconds. For 73.6% of patients, AI-FFR provided completely automatic lesion detection. For the remaining patients, semiautomated evaluations using manual lesion marking were required. The sensitivity of AI-FFR measurements was 90.2%. In addition, the specificity was 94.9%, positive predictive value was 83.5%, negative predictive value was 97.1%, overall diagnostic accuracy was 93.8% and the area under the ROC curve was 0.93.
AI-FFR measurements were also accurate for lesions with invasive FFR values in the “gray zone” range.
“In this retrospective, international, multicenter study, we evaluated the diagnostic performance of an automated AI-based software that calculates AI-FFR values directly from coronary angiograms without the need for pressure wires, adenosine, or specific angiographic acquisition,” wrote first author Eyal Ben-Assa, MD, with Assuta Ashdod University Hospital in Israel and the Massachusetts Institute of Technology in Cambridge, Massachusetts, and colleagues. “The principal findings are that AI-FFR demonstrated high diagnostic accuracy compared with invasive wire-based FFR across a wide range of patients and lesions.”
The group also pointed to certain limitations of this study. It was retrospective, for example, and the “meticulousness with which the invasive wire-based FFR technique was performed could not be verified." Also, the medtech company behind AutocathFFR, MedHub-AI, funded this analysis.
Even with those details in mind, however, the group pointed to the value of these findings.
“The present international, multicenter retrospective study demonstrates that AI-FFR—an automated, machine learning-based FFR tool—can achieve high diagnostic accuracy compared with invasive FFR, both within and outside the critical gray zone,” the group wrote. “Its rapid analysis time, minimal procedural requirements, and operator independence in most cases may support broader adoption of physiologic assessment during routine coronary angiography. Prospective studies are needed to evaluate its utility in guiding revascularization decisions and improving patient outcomes.”
Click here for the full study.
Michael has more than 19 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.