Fractional Flow Reserve Derived From Routine Coronary Angiograms

Pressure wire-based fractional flow reserve (FFR) has become the standard of reference for decision making regarding coronary revascularization. Deriving FFR from routine angiograms could facilitate the uptake of FFR-based clinical decisions. Several angiography-derived FFR methods have recently been introduced (1,2). These methods are based on computational fluid dynamic simulations. FFR_angio is a novel technology providing a functional angiography mapping of the coronaries in 3D based on a rapid flow analysis of a dynamically derived lumped model that can assess FFR using routine angiograms and hemodynamic data.

The primary element of the angiogram-based FFR measurement is the 3-dimensional (3D) rebuilding of the coronary tree, after which the system scans the entire reconstructed tree in 3D and analyzes each branch as well as each bifurcation (or trifurcation), looking for narrowed regions. A hemodynamic evaluation follows, where the contribution of each narrowing to the total resistance to flow is taken into account and a subsequent lumped model is built. This allows pressure drops and flow rates to be estimated. The accumulated volume of the coronary tree and the total coronary length, calculated from a reconstruction of its geometry, enable an estimation of normal supply through an assessment of the microcirculatory bed resistance. The solution of the lumped model based on the inlet and outlet boundary conditions allows us to evaluate ratios of flow rate for stenosed versus “healthy” coronary trees.

Eighty-eight patients with stable angina were included in this study. FFR measurements were performed for clinical reasons in ≥1 coronary artery. The stenosis was clearly delineated on the angiograms. Patients with left main stenoses, ostial stenosis, in-stent restenosis at the target vessel, and previous bypass surgery were excluded (7.7%).

At least 2 angiographic projections of the vessel to be measured were acquired at 15 frames/s. The exact inclination of the radiographic tube was left to the operator’s discretion. Care was taken to fill the artery as completely as possible with contrast medium and to image the entire coronary tree at each view. The FFR_angio computations were performed offline from Digital Imaging and Communications in Medicine (DICOM) format files by operators not present in the catheterization laboratory and blinded to the invasive FFR results. Each series of DICOM cine sequences was loaded and processed along with the patient’s mean aortic pressure obtained at the time the angiogram was acquired. User interaction was required to guide automatic processing and included verification of cardiac phase synchronization and proper extraction of vessel centerlines and radii. The automatic processing consisted of the 3D tree reconstruction and the flow simulation.

Invasive FFR measurements were performed in duplicate with 6-F guide catheters, a pressure monitoring wire, and intracoronary adenosine (100–200 μg). Care was taken to document the exact anatomical position of the sensor. To test interobserver variability and the possible influence of human factors on the results of FFR_angio, 2 independent operators analyzed all angiograms.

A total of 101 lesions were analyzed, 30% of which had FFR values between 0.70 and 0.90. FFR_angio was calculated and compared to invasive FFR measurements at the exact location of the sensor (Figure 1A). A high degree of concordance was found between 2 measurements of FFR_angio performed by 2 different operators (interclass correlation coefficient of 0.97; p < 0.001). Figure 1B shows linear regression analyses for the 2 independent observers. We found that invasive FFR was a robust predictor of FFR_angio, with 86.8% of the variability of FFR_angio explained by the invasive FFR values; the β coefficient of invasive FFR was 0.855 (95% confidence interval [CI]: 0.789 to 0.922) and the intercept was 0.124 (95% CI: 0.068 to 0.179). The estimated bias was 0.004 ± 0.042 and did not differ significantly from zero. The 95% limits of agreement were −0.1 to 0.1. We used 0.8 as the cutoff value for FFR_angio, and found that the sensitivity, specificity, and diagnostic accuracy were 88%, 98%, and 94%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Lesion in the Left Coronary Tree

A case demonstrating a significant lesion in the left coronary tree (A). Correlation between invasive FFR and FFR_angio is represented by 2 different operators (B). FFR = fractional flow reserve; FFR_angio = angiography-derived FFR.

This first-in-human study indicated high reproducibility and diagnostic accuracy of FFR_angio compared with invasive FFR. The data were obtained in patients with characteristics encountered in most percutaneous coronary intervention trials and included relatively well-delineated lesions associated with a large range of FFR values. This high diagnostic accuracy is not only related to extreme values as almost one third of lesions were within a 5% range adjacent to the cutoff value. If the short turnaround time needed to obtain the FFR_angio value and the diagnostic accuracy are confirmed in larger studies, FFR_angio may foster a wider adoption of FFR-based decision making for revascularization in patients with coronary artery disease.

• American College of Cardiology Foundation