Fractional Flow Reserve Computed from Noninvasive CT Angiography Data: Diagnostic Performance of an On-Site Clinician-operated Computational Fluid Dynamics Algorithm

Adriaan Coenen, Marisa Lubbers, Akira Kurata, A Kono, Admir Dedic, Raluca Chelu, Marcel Dijkshoorn, Frank Gijsen, Mohamed Ouhlous, Robert Jan van Geuns, Koen Nieman

Research output: Contribution to journalArticleAcademicpeer-review

216 Citations (Scopus)

Abstract

Purpose: To validate an on-site algorithm for computation of fractional flow reserve (FFR) from coronary computed tomographic (CT) angiography data against invasively measured FFR and to test its diagnostic performance as compared with that of coronary CT angiography. Materials and Methods: The institutional review board provided a waiver for this retrospective study. From coronary CT angiography data in 106 patients, FFR was computed at a local workstation by using a computational fluid dynamics algorithm. Invasive FFR measurement was performed in 189 vessels (80 of which had an FFR <= 0.80); these measurements were regarded as the reference standard. The diagnostic characteristics of coronary CT angiography-derived computational FFR, coronary CT angiography, and quantitative coronary angiography were evaluated against those of invasively measured FFR by using C statistics. Sensitivity and specificity were compared by using a two-sided McNemar test. Results: For computational FFR, sensitivity was 87.5% (95% confidence interval [CI]: 78.2%, 93.8%), specificity was 65.1% (95% CI: 55.4%, 74.0%), and accuracy was 74.6% (95% CI: 68.4%, 80.8%), as compared with the finding of lumen stenosis of 50% or greater at coronary CT angiography, for which sensitivity was 81.3% (95% CI: 71.0%, 89.1%), specificity was 37.6% (95% CI: 28.5%, 47.4%), and accuracy was 56.1% (95% CI: 49.0%, 63.2%). C statistics revealed a larger area under the receiver operating characteristic curve (AUC) for computational FFR (AUC, 0.83) than for coronary CT angiography (AUC, 0.64). For vessels with intermediate (25%69%) stenosis, the sensitivity of computational FFR was 87.3% (95% CI: 76.5%, 94.3%) and the specificity was 59.3% (95% CI: 47.8%, 70.1%). Conclusion: With use of a reduced-order algorithm, computation of the FFR from coronary CT angiography data can be performed locally, at a regular workstation. The diagnostic accuracy of coronary CT angiography-derived computational FFR for the detection of functionally important coronary artery disease (CAD) was good and was incremental to that of coronary CT angiography within a population with a high prevalence of CAD. (C)RNSA, 2014.
Original languageUndefined/Unknown
Pages (from-to)674-683
Number of pages10
JournalRadiology
Volume274
Issue number3
DOIs
Publication statusPublished - 2015

Research programs

  • EMC COEUR-09
  • EMC NIHES-03-30-01

Cite this