“Virtual” (Computed) Fractional Flow Reserve

Fractional flow reserve (FFR) is the “gold standard” for assessing the physiological significance of coronary artery disease during invasive coronary angiography. FFR-guided percutaneous coronary intervention improves patient outcomes and reduces stent insertion and cost; yet, due to several practical and operator related factors, it is used in <10% of percutaneous coronary intervention procedures. Virtual fractional flow reserve (vFFR) is computed using coronary imaging and computational fluid dynamics modeling. vFFR has emerged as an attractive alternative to invasive FFR by delivering physiological assessment without the factors that limit the invasive technique. vFFR may offer further diagnostic and planning benefits, including virtual pullback and virtual stenting facilities. However, there are key challenges that need to be overcome before vFFR can be translated into routine clinical practice. These span a spectrum of scientific, logistic, commercial, and political areas. The method used to generate 3-dimensional geometric arterial models (segmentation) and selection of appropriate, patient-specific boundary conditions represent the primary scientific limitations. Many conflicting priorities and design features must be carefully considered for vFFR models to be sufficiently accurate, fast, and intuitive for physicians to use. Consistency is needed in how accuracy is defined and reported. Furthermore, appropriate regulatory and industry standards need to be in place, and cohesive approaches to intellectual property management, reimbursement, and clinician training are required. Assuming successful development continues in these key areas, vFFR is likely to become a desirable tool in the functional assessment of coronary artery disease.

own visual assessment is physiologically accurate, allied to a misconception that multiple visual assessments (e.g., in a "Heart Team" setting) or a prior noninvasive test of ischemia improve their accuracy. Finally, despite the FAME (Fractional Flow Reserve Versus Angiography for Multivessel Evaluation) trial data (5,6), some clinicians remain skeptical of the value of PCI in the context of stable coronary artery disease (7), which reduces enthusiasm for invasive FFR assessment.

VIRTUAL FFR
There is, therefore, a need for a method that delivers the benefits of physiological lesion assessment to every cardiologist without the practical drawbacks that limit the inva- Arterial anatomy is "segmented" from coronary imaging (computed tomography [CT] or invasive angiography) and reconstructed into 2-or 3-dimensional (3D) in silico surface representations. These arterial models must then be discretized, or "meshed," into a finite number of volumetric elements. In addition, the time-step of the simulation must be defined. The processes of spatial and temporal "discretization" determines the fidelity and refinement of a given analysis. The physical conditions at the inlet, outlet(s), and arterial walls must then be defined (i.e., the "boundary conditions"). A computer file that fully defines the spatial, temporal, and physiological bounds of the modeled system is then generated and executed using specialist software known as a CFD "solver," which simulates the distribution and dynamics of blood pressure, flow, and shear stress within the artery over time. These data are used to generate predictions regarding pressure and flow changes across coronary stenoses, from which vFFR can be calculated at any point along the vessel. These processes and the key stages of developing a typical vFFR workflow are demonstrated in Figure 1. Steps) trial further assessed vFFR CT , utilizing "updated proprietary software," "improved segmentation," "refined physiological models," and "increased automation," which generated improved diagnostic accuracy in a larger cohort of 251 patients (484 vessels) (per-patient accuracy 81% (95% CI: 76% to 85%) (10). Subsequently, HeartFlow, Inc. has gained U.S.
Food and Drug Administration (FDA) approval for the use of vFFR CT as a class II Coronary Physiologic Simulation Software Device (11). vFFR CT is computed using standard CTCA protocols without induction of hyperemia and is, therefore, a highly practical and  Coronary angiogram (A) is "segmented" and reconstructed (B) into a 3-dimensional (3D) model (C). Surface and volumetric meshing "discretize" the patient-specific geometry (C). The physiological conditions beyond the modeled section must be represented at each boundary, that is, "boundary conditions" (D). Computational fluid dynamics simulation computes the pressure gradient, using the anatomical 3D model "tuned" with physiological parameters. Pressure ratio is computed from output data (E). Results are validated against invasive measurements during development (F). vFFR ¼ virtual fractional flow reserve.

ADVANTAGES OF vFFR
Early vFFR results using CT (16) and angiographic images (13,14) are encouraging, and in the case of  (Figure 2). It can also provide a "virtual stenting" facility, whereby the physiological effect of alternative interventional strategies can be trialed in silico (by computer simulation) before treatment is delivered in vivo. vFFR can also assess any segment of the coronary tree, including those to which it might be challenging to pass a pressure wire.

CHALLENGES
Several scientific, logistic, and commercial challenges must be overcome, however, before vFFR can enter routine clinical use.

SEGMENTATION
Segmentation from medical images, whether they be derived from CTCA or CAG, is crucial to the accuracy of CFD simulation. CT-based vFFR is apposite for truly noninvasive vFFR, but CTCA is mainly used to rule out coronary artery disease in those with low-to-medium pre-test probability of coronary artery disease, rather than for detailed planning of revascularization. In many patients, CTCA does not provide sufficient image quality for accurate segmentation, whether due to cardiac or respiratory movement, tachycardia or arrhythmia leading to a "stair-step" artefact, phase misregistration or blurring, or calcification leading to "blooming" or "streaking" (17). The noninvasive nature of CTCA also prevents the measurement of other physiological parameters that can be used to inform CFD simulation. Segmentation from CAG is also challenging.
Current protocols most commonly segment from just 2 projections, which may under-represent the full 3D anatomy. Software is used to correct for patient movement between acquisitions, but this is not entirely robust, and biplane CAG acquisition is not available in the majority of catheter laboratories.
Rotational coronary angiography can be successful because it offers the potential to use multiple views in the reconstruction, or at least the option to select 2 optimal projections, eliminating vessel overlap, foreshortening, and inadequate opacification, but is also not widely available. Methods such as intravascular ultrasound and (particularly) optical coherence tomography would add detailed anatomical accuracy,  If vFFR is to succeed, a strategy that provides personalized estimation of the distal boundary conditions, using noninvasive clinical data, is required.

CFD SIMULATION
The optimal method of CFD simulation is also yet to be determined. The coronary circulation is a

COMPUTATION TIME
Prolonged vFFR computation times, over many hours, have been a major concern and may limit vFFR applicability. This is less crucial for CT-FFR, but is more important for CAG-related methods where real-time, on-table results are required. Offline, remote, supercomputer simulation represents one solution, but this is less attractive than system acceleration that allows real-time processing within software integrated into local catheter-laboratory systems. The adoption of very high-powered computation locally is not easily accommodated within parsimonious health care systems. Steadystate or reduced-order modeling are further options, with CFD results being generated within 5 min, which is comparable to invasive measurement (14).
Unlike invasive FFR, vFFR can also evaluate several coronaries and lesions in a single analysis. Using steady-state or 1-dimensional modeling is attractive but may sacrifice accuracy. However, the challenge of system acceleration appears eminently achievable.

MODEL COMPLEXITY AND DESIGN
Complex models, requiring invasive measurements with prolonged run times, have improved accuracy, but physicians require simple, rapid systems of sufficient accuracy to inform treatment decisions.
Further work is needed to discern models that balance these needs optimally. One potential solution is a multitiered approach that delivers fast results but with wide confidence margins, reserving more complex modeling for more borderline cases where increased precision is required. Furthermore, is there an appetite for remote supercomputation (raising issues of transferring large confidential datasets outside of the hospital), or would physicians prefer to run analyses themselves using systems within their catheterization laboratory?

ACCURACY AND VALIDATION
Accuracy is absolutely key to vFFR's success. However, what constitutes accuracy is yet to be defined.   In the context of interventional cardiology, CFD modeling is a new technology that is potentially disruptive, especially to manufacturers of hardware that may become redundant if vFFR is successful.
Traditionally-minded manufacturers and physicians will need to embrace these new techniques and engage with academics and modelers to ensure that the best methods are adopted.
Medicare reimbursement has boosted FFR use in the United States. Now that CTCA-based vFFR has been approved by the FDA, a similar arrangement is needed.
There are currently no industry standards regarding accuracy, reliability, or validation. The FDA is addressing this through a benchmarking initiative that aims to advance the application of CFD technology within the regulatory context (20). A consistent, evidence-based strategy administered by experts is also needed in Europe. Assuming successful development continues in these key areas, vFFR is likely to become a desirable tool in the functional assessment of coronary artery disease.
CT derived vFFR is emerging as a useful tool for low /medium risk patients whereas more invasive vFFR applications are emerging as useful tools in higher risk patients undergoing invasive management.