Machine Learning-Based Computed Tomography-Derived Fractional Flow Reserve Predicts Need for Coronary Revascularisation Prior to Transcatheter Aortic Valve Implantation
- Author:
Kai Dick David LEUNG
1
;
Pan Pan NG
;
Boris Chun Kei CHOW
;
Keith Wan Hang CHIU
;
Neeraj Ramesh MAHBOOBANI
;
Yuet-Wong CHENG
;
Eric Chi Yuen WONG
;
Alan Ka Chun CHAN
;
Augus Shing Fung CHUI
;
Michael Kang-Yin LEE
;
Jonan Chun Yin LEE
Author Information
- Publication Type:ORIGINAL ARTICLE
- From: Cardiovascular Imaging Asia 2025;9(1):2-8
- CountryRepublic of Korea
- Language:English
-
Abstract:
Objective:Patients with severe symptomatic aortic stenosis are assessed for coronary artery disease (CAD) prior to transcatheter aortic valve implantation (TAVI) with treatment implications. Invasive coronary angiography (ICA) is the recommended modality but is associated with peri-procedural complications. Integrating machine learning (ML)-based computed tomography-derived fractional flow reserve (CT-FFR) into existing TAVI-planning CT protocol may aid exclusion of significant CAD and thus avoiding ICA in selected patients.
Materials and Methods:A single-center, retrospective study was conducted, 41 TAVI candidates with both TAVI-planning CT and ICA performed were analyzed. CT datasets were evaluated by a ML-based CT-FFR software. Beta-blocker and nitroglycerin were not administered in these patients. The primary outcome was to identify significant CAD. The diagnostic performance of CT-FFR was compared against ICA.
Results:On per-patient level, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy were 89%, 94%, 80%, 97% and 93%, respectively. On per-vessel level, the sensitivity, specificity, PPV, NPV and diagnostic accuracy were 75%, 94%, 67%, 96% and 92%, respectively. The area under the receiver operative characteristics curve per individual coronary vessels yielded overall 0.90 (95% confidence interval 85%–95%). ICA may be avoided in up to 80% of patients if CT-FFR results were negative.
Conclusion:ML-based CT-FFR can provide accurate screening capabilities for significant CAD thus avoiding ICA. Its integration to existing TAVI-planning CT is feasible with the potential of improving the safety and efficiency of pre-TAVI CAD assessment.
