Introduction
Coronary Artery Disease, Computational Hemodynamics, and the Clinical Case for Multiscale Simulation
Coronary artery disease kills more people worldwide than any other single condition, and the haemodynamic consequences of coronary stenosis — elevated pressure drop, disturbed wall shear stress, and turbulent energy dissipation — are now understood to be as important as anatomical severity in determining ischaemic risk. Fractional flow reserve, defined as the ratio of mean distal coronary pressure to mean aortic pressure during pharmacological hyperaemia, has emerged as the clinical gold standard for functional stenosis assessment since the landmark FAME trials established its superiority over anatomical guidance alone. Yet FFR measurement requires catheterisation, adenosine administration, and a pressure wire, all of which carry procedural risk and cost.
Computational fluid dynamics offers a rigorous, physics-based route to non-invasive FFR estimation. HeartFlow's FFRCT product demonstrated this at commercial scale, achieving diagnostic accuracy comparable to invasive FFR in prospective multicentre trials. Nevertheless, full 3D Navier–Stokes simulations of the entire coronary tree remain computationally demanding — a single patient simulation requiring several million finite-element degrees of freedom, resolved over multiple cardiac cycles, can require tens of CPU-hours on dedicated hardware.
The alternative — reduced-order modelling using one-dimensional wave-propagation equations or lumped-parameter (0D) windkessel analogues — trades spatial resolution for computational tractability. These models can capture global coronary dynamics but lack the spatial resolution to resolve complex recirculating flow structures, shear stress distributions, and turbulent transitions that arise immediately upstream and downstream of a stenosis.
The natural resolution of this impasse is multiscale coupling: a framework in which a high-fidelity 3D CFD model resolves the stenotic region in detail, while a surrounding reduced-order model provides physiologically appropriate boundary conditions. The present work addresses this gap by presenting a multiscale framework in which a 3D Navier–Stokes simulation of the stenotic coronary segment is bidirectionally coupled to a 1D–0D vascular network model, with rigorous mathematical formulation of the coupling interface, patient-specific calibration, and clinical validation against measured FFR.
Advanced Literature Review and Critical Synthesis
Historical Trajectory, Governing Physics, Turbulence, Boundary Conditions, and Identified Contradictions
The application of CFD to coronary haemodynamics dates to the early 1990s, when Friedman and colleagues employed finite-element methods to study flow in idealized coronary bifurcation models. The clinical ambition to compute FFR non-invasively crystallised around 2012 with the publication of the DISCOVER-FLOW and DeFACTO trials, reporting area-under-curve values of 0.90–0.93 against invasive FFR using the HeartFlow FFRCT platform.
All major CFD studies of coronary haemodynamics are grounded in the incompressible Navier–Stokes equations. A substantial fraction of the literature adopts the Newtonian assumption, treating blood viscosity as a constant $\mu \approx 0.003$–$0.004\ \text{Pa·s}$. Ballyk and colleagues demonstrated that non-Newtonian effects modelled via the Carreau constitutive equation alter WSS magnitudes by up to 20% in recirculation zones, specifically in regions where shear rates drop below $10\ \text{s}^{-1}$.
For a 70% area-stenosis in a typical left anterior descending (LAD) artery with a resting flow of ~1 mL/s, the peak Reynolds number in the stenotic throat exceeds 2500 — above the classical transition threshold for pipe flow. Varghese and Frankel demonstrated using direct numerical simulation that turbulent fluctuations arise in the poststenotic jet for stenoses above 50% diameter reduction, and that the k–ω SST formulation provides acceptable predictions of time-averaged pressure drop.
The Windkessel outlet model represents the downstream vasculature as a lumped two- or three-element RC electrical analogue. More sophisticated outlet models including structured trees derived from fractal coronary microvascular networks and coupled 1D coronary wave-propagation models enable bidirectional coupling between the 3D domain and a 1D network. This is precisely the strategy adopted in the present framework.
Several substantive disagreements within the literature warrant explicit identification. The table below summarises the major contradictions, their sources, and resolution strategies adopted here.
| Contradiction | Position A | Position B | Resolution Strategy |
|---|---|---|---|
| Newtonian vs. non-Newtonian rheology | Newtonian adequate for FFR (Taylor et al., 2013) | Non-Newtonian affects WSS ≥20% in recirculation (Ballyk et al., 1994) | Carreau–Yasuda adopted; Newtonian sensitivity tested |
| Rigid vs. compliant walls | Rigid wall acceptable for pressure drop | FSI alters WSS up to 15% in systole (Torii et al., 2009) | Rigid wall with FSI sensitivity study in future work |
| Laminar vs. turbulent flow | Laminar valid for <75% stenoses | Turbulent bursts occur above 50% stenosis (Varghese & Frankel) | k–ω SST with LRN correction adopted |
| Windkessel vs. 1D outlet BC | Windkessel sufficient for FFR (HeartFlow) | 1D coupling captures wave reflections affecting FFR (Blanco et al.) | 1D network coupling is the central methodological contribution |
| Steady vs. pulsatile simulation | Steady hyperaemia adequate for FFR | Pulsatile simulation necessary for WSS and instability | Pulsatile simulation adopted throughout |
| Generic vs. patient-specific resistance | Murray's law scaling sufficient (FFRCT) | Patient-specific calibration needed (Nørgaard et al.) | Calibration from CT-derived myocardial volume |
Table 1. Major methodological contradictions in coronary haemodynamics CFD.
Research Gaps, Novelty Statement, and Hypothesis
Systematic analysis of the existing literature reveals five high-impact unresolved problems. First, existing coronary 3D–1D frameworks employ one-way or weakly coupled approaches without iterative convergence guarantees. Second, network resistance parameters are drawn from population averages without patient-specific uncertainty quantification. Third, the interaction between cardiac pulsatility and stenosis-induced turbulent transition is poorly characterised. Fourth, formal uncertainty quantification studies for WSS under geometric and boundary condition variability are sparse. Fifth, the specific combination of 3D–1D coupling, patient-specific calibration, and prospective FFR validation has not been demonstrated.
The present study makes four original contributions: (i) a bidirectionally coupled 3D/1D–0D framework with Robin-stabilised iterative interface guaranteeing convergence; (ii) a patient-specific network calibration methodology with explicit uncertainty propagation; (iii) Carreau–Yasuda non-Newtonian rheology in the 3D domain coupled to a Newtonian 1D network; and (iv) prospective validation against invasive FFR.
A multiscale framework coupling a localised 3D Navier–Stokes simulation of the stenotic coronary segment to a patient-calibrated 1D–0D coronary network model will predict fractional flow reserve with a mean absolute error of ≤0.03 FFR units against invasive pressure-wire measurements, while reducing computational cost by ≥60% compared to a full 3D coronary simulation, without sacrificing the spatial resolution of critical haemodynamic quantities in the stenotic zone.
Mathematical Formulation
Domain Decomposition, Governing Equations, Rheology, Network Model, and Coupling Strategy
Let $\Omega$ denote the total spatial domain of the coronary circulation. We partition $\Omega$ into two non-overlapping subdomains: $\Omega_{3D}$, a three-dimensional region enclosing the stenotic segment, and $\Omega_{1D}$, the complementary one-dimensional network representing the proximal and distal coronary vessels. The interface $\Gamma = \partial\Omega_{3D} \cap \partial\Omega_{1D}$ consists of inlet and outlet cross-sections at which coupling conditions enforce continuity of flow rate and pressure.
Within $\Omega_{3D}$, conservation of mass for an incompressible fluid is expressed as:
where $\mathbf{u} = \mathbf{u}(\mathbf{x}, t)$ is the velocity field and $T$ is the total simulation time. Blood density $\rho = 1060\ \text{kg/m}^3$ is assumed constant throughout.
The full incompressible Navier–Stokes momentum equation in the 3D domain reads:
where $p$ is the mechanical pressure, $\boldsymbol{\tau}$ is the viscous stress tensor, and $\mathbf{f}$ represents body forces (negligible for coronary flow). For a generalised Newtonian fluid:
where $\dot{\gamma} = \sqrt{2\,\mathbf{D}:\mathbf{D}}$ is the scalar shear rate. The viscosity function $\mu(\dot{\gamma})$ follows the Carreau–Yasuda model:
with parameters fitted to human blood: $\mu_0 = 0.056\ \text{Pa·s}$, $\mu_\infty = 0.00345\ \text{Pa·s}$, $\lambda = 3.313\ \text{s}$, $n = 0.3568$, and $a = 2.0$. In regions of high shear ($\dot{\gamma} \gg \lambda^{-1}$), $\mu \to \mu_\infty$ and the model reduces to the Newtonian case.
Within $\Omega_{1D}$, blood flow in each vessel segment of length $L$ and cross-sectional area $A(x,t)$ is governed by the one-dimensional conservation equations:
where $U$ is the cross-sectional-mean velocity, $\alpha$ is the momentum-flux correction factor ($\alpha = 4/3$ for Poiseuille flow), and $K_r$ is the viscous resistance term per unit length. For rigid 1D segments this simplifies to $\partial Q/\partial x = 0$ where $Q = AU$.
The tube law relating transmural pressure to cross-sectional area provides the closure relation for compliant segments:
where $A_0$ is the reference area, $p_\text{ext}$ is the external (intramyocardial) pressure, and $\beta$ is a wall stiffness parameter. At each terminal node, a three-element Windkessel (RCR) model represents the microvascular bed:
where $R_t = R_p + R_d$ is the total terminal resistance, $C$ is the microvascular compliance, and $p_\text{ven} = 5\ \text{mmHg}$. The partition $R_p/R_d = 0.06/0.94$ follows Stergiopulos et al.
The coupled problem is solved via a partitioned iterative scheme. At the interface $\Gamma$, the coupling conditions are:
To stabilise the iteration against the classical Dirichlet–Neumann divergence in haemodynamic applications, we employ Robin interface conditions:
where $\gamma > 0$ is a coupling stabilisation parameter (units: vascular resistance) set to the characteristic impedance of the vessel at the coupling interface. Convergence is declared when both relative residuals fall below $\varepsilon = 10^{-4}$. The system is integrated using an implicit second-order backward differencing (BDF2) scheme in the 3D domain and an explicit Adams–Bashforth predictor in the 1D network, with time step $\Delta t = 1\ \text{ms}$.
The wall shear stress vector on the coronary artery wall $\partial\Omega_w$ is defined as:
where $\hat{n}$ is the outward unit normal to $\partial\Omega_w$. Time-averaged WSS magnitude (TAWSS) and oscillatory shear index (OSI) are computed over one cardiac cycle $T$:
Fractional flow reserve is computed as:
where $\bar{p}_d$ is the time-averaged distal coronary pressure and $\bar{p}_a$ is the time-averaged aortic pressure, both evaluated at simulated pharmacological hyperaemia.
For cases where transition to turbulence is anticipated (peak Reynolds number $\text{Re} > \text{Re}_\text{crit} = 2000$), the RANS equations are closed using the k–ω Shear Stress Transport (SST) model:
where $k$ is the turbulent kinetic energy, $\omega$ is the specific dissipation rate, $\mu_t = \rho k/\max(\omega,\, \Omega F_2/a_1)$ is the turbulent viscosity, $F_1$ and $F_2$ are blending functions, and model constants take the SST values recommended by Menter. Low-Reynolds-number damping is applied through the $R_y$-based modification of the $\beta^*$ coefficient.
Methodology
Patient Cohort, Geometry Reconstruction, Network Calibration, Mesh Generation, and Solver Configuration
The validation dataset comprises patients with stable angina and angiographically identified intermediate coronary stenoses (40–80% diameter reduction) referred for invasive physiological assessment. Inclusion criteria require: (i) high-resolution coronary CTA (≥64-slice, prospective ECG gating, resolution ≤0.6 × 0.6 × 0.6 mm³); (ii) invasive FFR measurement by pressure wire during adenosine-mediated hyperaemia; (iii) absence of prior coronary intervention in the target vessel; and (iv) left ventricular ejection fraction ≥ 45%.
Coronary lumen segmentation is performed from CTA DICOM datasets using a semi-automated pipeline combining multi-threshold active contour initialisation with level-set evolution. Surface meshes are generated via marching-cubes algorithm with Laplacian smoothing (20 iterations, coefficient 0.2), calibrated to preserve the stenosis minimum lumen area within ±3% of the pre-smoothing value. The 3D domain spans the stenotic segment plus 3 proximal and 3 distal diameters.
Myocardial mass $M_\text{myo}$ is estimated from CTA volume using Cavalieri's method. Total coronary blood flow at rest is estimated from the allometric relationship:
Terminal resistance at each outlet is:
evaluated at resting mean aortic pressure ($\bar{p}_a = 90\ \text{mmHg}$). Under simulated hyperaemia, $R_t$ is scaled by $f_\text{hyp} = 0.24$ (four-fold flow increase). Sensitivity to ±20% variation in $f_\text{hyp}$ is assessed in the uncertainty analysis.
The 3D computational mesh is generated using ANSYS ICEM-CFD with a hybrid strategy: structured prismatic boundary layers (5 layers, expansion ratio 1.2, first-layer height 15 µm, $y^+ \approx 0.5$–$1.0$) and an unstructured polyhedral core. Mesh independence is declared when the change in predicted FFR between successive refinement levels is less than 0.005 units.
At the inlet interface $\Gamma_\text{in}$, the time-varying velocity profile is prescribed as the Womersley solution corresponding to the instantaneous flow rate $Q(t)$ delivered by the 1D model, with Womersley parameter:
At the outlet interface $\Gamma_\text{out}$, the instantaneous pressure from the 1D model is applied as a Neumann boundary condition. The coronary artery wall is modelled as rigid and no-slip. Flow is initialised from the steady Stokes solution and run for three cardiac cycles before extracting cycle-averaged quantities.
The 3D Navier–Stokes equations are solved using finite-volume discretisation in OpenFOAM (version 10) with the PISO algorithm and two corrector steps per time step. Convective terms use a second-order linear upwind scheme with van Leer flux limiting. The 1D network employs a second-order Taylor–Galerkin finite-element scheme with element length $h_{1D} = 1\ \text{mm}$. The 0D Windkessel equations are integrated with an explicit fourth-order Runge–Kutta scheme. Communication between 3D and 1D solvers is implemented via an in-memory MPI interface. Simulations are run on a 64-core workstation (Intel Xeon Platinum 8380, 2.3 GHz, 512 GB RAM).
Validation Framework and Error Analysis
The primary validation metric compares $\text{FFR}_\text{CFD}$ against invasively measured $\text{FFR}_\text{inv}$. Agreement is quantified by:
alongside Pearson correlation coefficient $r$ and Bland–Altman analysis. Diagnostic performance uses the clinical threshold FFR ≤ 0.80, with sensitivity, specificity, PPV, NPV, and AUC-ROC reported.
A global sensitivity analysis using Sobol' variance-based indices is performed from a Monte Carlo ensemble of $N = 512$ coupled simulations varying: stenosis minimum lumen diameter (±0.3 mm), hyperaemic resistance reduction factor $f_\text{hyp}$ (±20%), inlet flow rate (±15%), blood viscosity $\mu_\infty$ (±10%), and aortic pressure (±8 mmHg). The dominant source of FFR uncertainty is anticipated to be $f_\text{hyp}$, with stenosis geometry as second-ranked contributor.
Wall-clock computation time is compared between the proposed coupled framework (3D domain restricted to the stenotic segment) and a standalone full-domain 3D simulation. The mesh element count is reduced from approximately 8–12 million elements (full coronary tree) to 500,000–2,000,000 elements (stenotic segment only), with an anticipated wall-clock speedup of 4–8× (from 48–96 hours to 8–16 hours on the same 64-core hardware).
Proposed Results and Interpretation
The multiscale framework is expected to reveal a characteristic haemodynamic signature in the stenotic region: a high-velocity jet in the stenotic throat with peak velocities of 2–4 m/s under hyperaemia, elevated TAWSS (>10 Pa) on the proximal shoulder of the plaque, and a low-WSS, high-OSI region in the recirculation zone distal to the stenosis. OSI > 0.3 identifies sites of disturbed, oscillatory shear stress that promote inflammatory gene expression.
Based on published performance benchmarks and the methodological improvements of the present framework, we anticipate: $\text{MAE} \approx 0.025$–$0.035$ FFR units, Pearson correlation $r \geq 0.90$, and $\text{AUC-ROC} \geq 0.90$. The Bland–Altman analysis is expected to reveal a small systematic bias toward CFD overestimation, consistent with the tendency of simplified boundary conditions to underestimate distal microvascular resistance.
The restriction of the 3D domain to the stenotic segment reduces mesh element count by approximately 75–95% relative to full-tree 3D simulation, with the 1D network adding negligible overhead. The anticipated 60–75% reduction in wall-clock time is consistent with reported speedups in analogous multiscale frameworks applied to cerebrovascular and aortic problems.
Clinical Significance
The clinical motivation for non-invasive FFR computation is well established. The multiscale framework addresses two specific barriers to clinical deployment of existing CFD-FFR platforms. First, by reducing computational cost relative to full-domain 3D simulation, the framework moves toward the turnaround times (<2 hours) necessary for integration into routine clinical workflows. The 8–16 hour estimate remains above this threshold but represents a meaningful step; further acceleration through GPU-parallel 3D solvers or surrogate model hybrid approaches is tractable.
Second, the explicit representation of downstream coronary disease within the 1D network enables more accurate FFR prediction in patients with multivessel or diffuse disease. Beyond FFR, the wall shear stress maps resolved at sub-millimetre scale supplement anatomical and physiological stenosis assessment with mechanobiological information relevant to long-term disease management. Low TAWSS (<0.4 Pa) at the downstream edge of the stenosis is mechanistically associated with endothelial activation and plaque progression; high OSI (>0.3) identifies sites of disturbed shear promoting inflammatory gene expression.
Limitations
The framework rests on several assumptions that constrain its current scope. The rigid-wall approximation neglects cyclic coronary deformation during the cardiac cycle, which can alter near-wall velocity profiles and WSS by 10–20% during systole. Coronary autoregulation is represented only through a static resistance parameter, appropriate for steady hyperaemia but limiting accuracy in dynamic exercise testing or post-intervention simulations. The 1D network model assumes vessels operate in the linear haemodynamic regime, which may not hold under severe coronary hypertension or vasospasm. Finally, the validation dataset size limits the power of subgroup analyses; a larger prospective validation study is required before regulatory consideration.
Future Directions
Several extensions are identified as high-priority research directions. Fluid–structure interaction with patient-specific vessel wall properties estimated from 4D CTA or intravascular elastography would address the rigid-wall limitation. Coupling to a coronary autoregulation model based on metabolic feedback would enable simulation of physiological responses to exercise, post-stent hyperaemia, or pharmacological interventions other than adenosine.
Physics-informed neural networks trained on the CFD ensemble generated in the uncertainty analysis offer a promising route to near-real-time FFR prediction. The PINN architecture could encode the Navier–Stokes equations as soft constraints, improving generalisation while retaining physical interpretability. Integration with coronary CT-derived FFR platforms, with the multiscale framework providing improved boundary conditions for commercial solvers, is a near-term translational goal.
Methodological Comparison with Representative Prior Work
Structured Comparison Matrix of Coronary CFD Studies
| Study | Geometry | Rheology | Turbulence | Outlet BC | Validation | Multiscale |
|---|---|---|---|---|---|---|
| Taylor et al. 2013 (FFRCT) | Patient-specific CTA | Newtonian | Laminar | Windkessel RCR | DISCOVER-FLOW | No (0D only) |
| Krams et al. 2005 | Patient-specific IVUS | Newtonian | Laminar | Zero stress | Histology | No |
| Ballyk et al. 1994 | Idealised bifurcation | Carreau | Laminar | Parabolic outlet | None | No |
| Varghese & Frankel 2003 | Idealised stenosis | Newtonian | DNS/RANS | Periodic | DNS benchmark | No |
| Blanco & Feijóo 2013 | Idealised + 1D network | Newtonian | Laminar | 1D coupling | Synthetic | 3D–1D |
| Kim et al. 2010 | Patient-specific CTA | Newtonian | Laminar | 0D heart model | Limited clinical | 3D–0D |
| Mynard et al. 2015 | 1D coronary network | Newtonian | N/A | 0D Windkessel | Pressure measurements | 1D–0D |
| Present Framework | Patient-specific CTA (3D+1D) | Carreau–Yasuda (3D), Newtonian (1D) | k–ω SST (LRN) | 1D–0D coupled | Invasive FFR | 3D–1D–0D |
Table 2. Methodological comparison matrix of coronary CFD studies.
- [1]Taylor CA, Fonte TA, Min JK. Computational fluid dynamics applied to cardiac computed tomography for noninvasive quantification of fractional flow reserve. J Am Coll Cardiol. 2013;61(22):2233–41.
- [2]Krams R, Wentzel JJ, Oomen JAF, et al. Evaluation of endothelial shear stress and 3D geometry as factors determining the development of atherosclerosis in human coronary arteries. Arterioscler Thromb Vasc Biol. 1997;17(10):2061–5.
- [3]Formaggia L, Gerbeau JF, Nobile F, Quarteroni A. On the coupling of 3D and 1D Navier–Stokes equations for flow problems in compliant vessels. Comput Methods Appl Mech Eng. 2001;191(6–7):561–82.
- [4]Quarteroni A, Veneziani A. Analysis of a geometrical multiscale model based on the coupling of ODEs and PDEs for blood flow simulations. SIAM J Multiscale Model Simul. 2003;1(2):173–95.
- [5]Vignon-Clementel IE, Figueroa CA, Jansen KE, Taylor CA. Outflow boundary conditions for three-dimensional finite element modeling of blood flow and pressure in arteries. Comput Methods Appl Mech Eng. 2006;195(29–32):3776–96.
- [6]Varghese SS, Frankel SH. Numerical modeling of pulsatile turbulent flow in stenotic vessels. J Biomech Eng. 2003;125(4):445–60.
- [7]Ballyk PD, Steinman DA, Ethier CR. Simulation of non-Newtonian blood flow in an end-to-end anastomosis. Biorheology. 1994;31(5):565–86.
- [8]Blanco PJ, Feijóo RA. A dimensionally-heterogeneous closed-loop model for the cardiovascular system and its applications. Med Eng Phys. 2013;35(5):652–67.
- [9]Mynard JP, Nithiarasu P. A 1D arterial blood flow model incorporating ventricular pressure, aortic valve and regional coronary flow using the LCG method. Commun Numer Methods Eng. 2008;24(5):367–417.
- [10]Kim HJ, Vignon-Clementel IE, Coogan JS, et al. Patient-specific modeling of blood flow and pressure in human coronary arteries. Ann Biomed Eng. 2010;38(10):3195–209.
- [11]Menter FR. Two-equation eddy-viscosity turbulence models for engineering applications. AIAA J. 1994;32(8):1598–605.
- [12]Mynard JP, Smolich JJ. One-dimensional haemodynamic modeling and wave dynamics in the entire adult circulation. Ann Biomed Eng. 2015;43(6):1443–60.
- [13]Nørgaard BL, Leipsic J, Gaur S, et al. Diagnostic performance of noninvasive fractional flow reserve derived from coronary computed tomography angiography. J Am Coll Cardiol. 2014;63(12):1145–55.
- [14]Torii R, Oshima M, Kobayashi T, et al. Fluid–structure interaction modeling of blood flow and cerebral aneurysm. Comput Methods Appl Mech Eng. 2009;198(45–46):3613–21.
- [15]Stergiopulos N, Young DF, Rogge TR. Computer simulation of arterial flow with applications to arterial and aortic stenoses. J Biomech. 1992;25(12):1477–88.
- [16]Eck VG, Donders WP, Sturdy J, et al. A guide to uncertainty quantification and sensitivity analysis for cardiovascular applications. Int J Numer Methods Biomed Eng. 2016;32(8):e02755.
- [17]Badia S, Nobile F, Vergara C. Fluid–structure partitioned procedures based on Robin transmission conditions. J Comput Phys. 2008;227(14):7027–51.
- [18]Sobol' IM. Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates. Math Comput Simul. 2001;55(1–3):271–80.
- [19]Antiga L, Piccinelli M, Botti L, et al. An image-based modeling framework for patient-specific computational hemodynamics. Med Biol Eng Comput. 2008;46(11):1097–112.
- [20]Lesage D, Angelini ED, Bloch I, Funka-Lea G. A review of 3D vessel lumen segmentation techniques. Med Image Anal. 2009;13(6):819–45.
- [21]Tröbs M, Achenbach S, Röther J, et al. Comparison of fractional flow reserve based on computational fluid dynamics modeling versus invasively measured fractional flow reserve. Am J Cardiol. 2016;117(1):29–35.
- [22]Gould KL, Johnson NP, Bateman TM, et al. Anatomic versus physiologic assessment of coronary artery disease. J Am Coll Cardiol. 2013;62(18):1639–53.
- [23]De Bruyne B, Pijls NHJ, Kalesan B, et al. Fractional flow reserve-guided PCI versus medical therapy in stable coronary disease. N Engl J Med. 2012;367(11):991–1001.
- [24]Sughimoto K, Lau YM, Lim BHP, et al. Computational hemodynamics of coronary artery stenosis: effect of varying degrees of stenosis on blood flow. J Med Biol Eng. 2022;42(5):558–70.
- [25]Pirola S, Cheng Z, Jarral OA, et al. On the choice of outlet boundary conditions for patient-specific analysis of aortic flow using CFD. J Biomech. 2017;60:15–21.
- [26]Murray CD. The physiological principle of minimum work: the vascular system and the cost of blood volume. Proc Natl Acad Sci. 1926;12(3):207–14.
- [27]Womersley JR. Method for the calculation of velocity, rate of flow and viscous drag in arteries when the pressure gradient is known. J Physiol. 1955;127(3):553–63.
- [28]Pedley TJ. The Fluid Mechanics of Large Blood Vessels. Cambridge University Press; 1980.
- [29]Ku DN. Blood flow in arteries. Annu Rev Fluid Mech. 1997;29:399–434.
- [30]Feijóo RA, Blanco PJ. Modelos Multiscala para el Sistema Cardiovascular. LNCC/MCTI; 2011.
- [31]Shim EB, Kamm RD, Heldt T, Mark RG. Numerical analysis of blood flow through a stenosed artery using a coupled, multiscale simulation method. Comput Cardiol. 2000;27:219–22.
- [32]Huo Y, Kassab GS. Hemodynamics in coronary arterial tree of serial stenoses. PLoS ONE. 2016;11(9):e0163715.
- [33]Xie S, Liu F, Duan Z, et al. A new CFD-based non-invasive method for functional diagnosis of coronary stenosis. Biomed Eng Online. 2018;17:36.
- [34]Ali S, Ho CY, Yang CC, et al. Computational fluid dynamics modeling of coronary artery blood flow using OpenFOAM: validation with the FDA benchmark nozzle model. J X-Ray Sci Technol. 2024;32(4).
- [35]Guo W, He W, Lu Y, et al. CT-FFR by expanding coronary tree with Newton-Krylov-Schwarz method. Eur Heart J Imaging Methods Pract. 2024;2(3):qyae106.
- [36]Maurya CS, Kumar A. Modelling of coronary artery stenosis and study of hemodynamic under the influence of magnetic fields. Comput Biol Med. 2025;184:109464.
- [37]Smits PC, et al. Deep operator learning for blood flow modelling in stenosed vessels. npj Artif Intell. 2025;1:35.
- [38]Marsden AL, et al. Personalized coronary and myocardial blood flow models incorporating CT perfusion imaging and synthetic vascular trees. npj Imaging. 2024;2:9.
- [39]Pijls NHJ, van Son JAM, Kirkeeide RL, et al. Experimental basis of determining maximum coronary, myocardial, and collateral blood flow by pressure measurements for assessing functional stenosis severity. Circulation. 1993;87(4):1354–67.
- [40]Tonino PAL, De Bruyne B, Pijls NHJ, et al. Fractional flow reserve versus angiography for guiding percutaneous coronary intervention (FAME). N Engl J Med. 2009;360(3):213–24.
- [41]Koo BK, Erglis A, Doh JH, et al. Diagnosis of ischemia-causing coronary stenoses by noninvasive FFR computed from coronary CTA (DISCOVER-FLOW). J Am Coll Cardiol. 2011;58(19):1989–97.
- [42]Steinman DA, Taylor CA. Flow imaging and computing: large artery hemodynamics. Ann Biomed Eng. 2005;33(12):1704–9.
- [43]Mulder G, Bogaerds ACB, Rongen P, van de Vosse FN. On automated Windkessel parameter estimation in patient-specific hemodynamic modeling. Int J Numer Methods Biomed Eng. 2011;27(3):313–24.
- [44]Morbiducci U, Ponzini R, Rizzo G, et al. Mechanistic insight into the physiological relevance of helical blood flow in the human aorta. Biomech Model Mechanobiol. 2011;10(3):339–55.
- [45]Pijls NHJ, Fearon WF, Tonino PAL, et al. Fractional flow reserve versus angiography for guiding PCI in patients with multivessel CAD: 2-year follow-up of the FAME study. J Am Coll Cardiol. 2010;56(3):177–84.
- [46]Müller LO, Toro EF. A global multiscale mathematical model for the human circulation with emphasis on the venous system. Int J Numer Methods Biomed Eng. 2014;30(7):681–725.
- [47]Gallo D, De Santis G, Negri F, et al. On the use of in vivo measured flow rates as boundary conditions for image-based hemodynamic models of the human aorta. Ann Biomed Eng. 2012;40(3):729–41.
- [48]Itu L, Rapaka S, Passerini T, et al. A machine-learning approach for computation of fractional flow reserve from coronary computed tomography. J Appl Physiol. 2016;121(1):42–52.
- [49]Zhong L, Zhang JM, Su B, et al. Application of patient-specific CFD in coronary artery disease: systematic review and limitations. Circ Cardiovasc Imaging. 2018;11(11):e005918.
- [50]Fearon WF, Bornschein B, Tonino PAL, et al. Economic evaluation of fractional flow reserve-guided PCI in patients with multivessel disease. Circulation. 2010;122(24):2545–50.
- [51]Wentzel JJ, Gijsen FJH, Stergiopulos N, et al. Shear stress, vascular remodeling and neointimal formation. J Biomech. 2003;36(5):681–8.
- [52]Perktold K, Rappitsch G. Computer simulation of local blood flow and vessel mechanics in a compliant carotid artery bifurcation model. J Biomech. 1995;28(7):845–56.
- [53]Quarteroni A, Ragni S, Veneziani A. Coupling between lumped and distributed models for blood flow problems. Comput Vis Sci. 2001;4(2):111–24.
- [54]Young DF, Tsai FY. Flow characteristics in models of arterial stenoses — I. Steady flow. J Biomech. 1973;6(4):395–410.
- [55]Johnston BM, Johnston PR, Corney S, Kilpatrick D. Non-Newtonian blood flow in human right coronary arteries: steady state simulations. J Biomech. 2006;39(6):1116–28.
- [56]Updegrove A, Wilson NM, Merkow J, et al. SimVascular: an open source pipeline for cardiovascular simulation. Ann Biomed Eng. 2017;45(3):525–41.
- [57]Razavi SE, Vassoury H, Madani S. Three-dimensional hemodynamic analysis of stenosis effects in coronary arteries using Navier–Stokes equation. Biomech Syst Technol. 2011;3:147–62.
- [58]Chiastra C, Morlacchi S, Gallo D, et al. Computational fluid dynamic simulations of image-based stented coronary bifurcation models. J R Soc Interface. 2013;10(84):20130193.
- [59]Xu P, Liu X, Zhang H, et al. Assessment of boundary conditions for CFD simulation in human carotid artery. Biomed Eng Online. 2020;17(1):168.
- [60]Douglas PS, Pontone G, Hlatky MA, et al. Clinical outcomes of fractional flow reserve by computed tomographic angiography-guided diagnostic strategies vs. usual care (PLATFORM). Eur Heart J. 2015;36(47):3359–67.