Beyond anatomy: individual physiology-enhanced approach to improve accuracy of fractional flow reserve from computed tomography
M Kruk, Z Ostrowski, A Kosior, M Rzycki, A Podkowka-Pozniak, C KepkaAbstract
Background
Fractional flow reserve derived from coronary computed tomography angiography (FFR-CT) is recommended for non-invasive evaluation of intermediate coronary artery stenoses. However, a recent meta-analysis showed that contemporary anatomy-only FFR-CT methods provide moderate agreement with invasive FFR with reduced discrimination in the intermediate “gray-zone” where clinical decisions are often most challenging. We hypothesized that integrating individual pressure-waveform hemodynamics into FFR-CT computation could improve functional assessment of coronary lesions, particularly in this diagnostically challenging range.
Methods
We developed a waveform-enhanced FFR-CT algorithm that incorporates Continuous Non-invasively measured Blood Pressure (CNBP) waveforms as patient-specific physiological to determine boundary conditions for Computational Fluid Dynamics models of coronary flow. FFR-CT values were computed along the coronary tree and adjusted to the invasive FFR measurements, with FFR ≤0.80 considered hemodynamically significant. In this multicenter study, 167 vessels from 132 patients enrolled in the NCT04777513 trial were analyzed. Diagnostic performance of waveform-enhanced FFR-CT was compared against invasive FFR and benchmarked against collected off–site FFR-CT results from a recent temporary meta-analysis including 2423 vessels, restricted to prospective diagnostic accuracy studies reporting vessel-level FFR-CT versus invasive FFR.
Results
Per-vessel analysis showed excellent diagnostic performance of waveform-enhanced FFR-CT, with an AUC of 0.916 (95% CI 0.868–0.956) and overall diagnostic accuracy of 88.6% for detecting functionally significant stenoses versus invasive FFR. Sensitivity and specificity were 92.5% and 87.4%, respectively, compared with 81.5% and 77.7% for benchmark FFR-CT (p=0.027 and p=0.0021), and AUC was significantly higher (0.916 vs 0.858; p=0.026). Bland–Altman analysis demonstrated minimal bias between invasive FFR and waveform-enhanced FFR-CT (mean difference −0.01, 95% limits of agreement −0.21 to 0.19) and a Pearson correlation coefficient of 0.69 with a regression slope close to unity. In the clinically relevant grey-zone subgroup (FFR 0.75–0.85), diagnostic accuracy remained high with waveform-enhanced FFR-CT (81.2% vs 65.4% for benchmark FFR-CT; p=0.003), with higher specificity (79.6% vs 59.7%; p=0.0014) and numerically higher sensitivity (85.0% vs 72.7%; p=0.14).
Conclusion
An individual physiology-enhanced FFR-CT approach that integrates patient-specific pressure–waveform dynamics into coronary CFD modeling improves diagnostic accuracy and agreement versus benchmark anatomy-only FFR-CT, including within the grey zone of intermediate lesions. These findings support the use of physiology-integrated FFR-CT to refine non-invasive selection of patients for invasive coronary angiography and revascularization.Accuracy of Cardiolens FFR-CT Pro