Journal of Biomechanics 161 (2023) 111857

Contents lists available at ScienceDirect

Journal of Biomechanics
journal homepage: www.elsevier.com/locate/jbiomech

The influence of flow distribution strategy for the quantification of

pressure- and wall shear stress-derived parameters in the coronary artery: A
CTA-based computational fluid dynamics analysis
Yibing Shi a, 1, Jin Zheng b, 1, Ying Zhang a, Quanlin Sun b, c, Jinhua Shen c, Yongguang Gao a,
Jingxi Sun a, Ning Yang a, Xuanxuan Zhou a, Suqing Li a, Jonathan R. Weir-McCall b, d,
Ping Xia a, *, Zhongzhao Teng b, c, *
a
Department of Radiology, Xuzhou Central Hospital, Xuzhou Clinical School of Xuzhou Medical University, Jiangsu, China
b
Department of Radiology, University of Cambridge, UK
c
Nanjing Jingsan Medical Science and Technology, Ltd., Jiangsu, China
d
Department of Radiology, Royal Papworth Hospital, Cambridge, UK

A R T I C L E I N F O A B S T R A C T

Keywords: For image-based computational fluid dynamics (CFD) analysis to characterize the local coronary hemodynamic
Hemodynamics environment, the accuracy depends on the flow rate which is in turn associated with outlet branches’
Coronary artery morphology. A good flow distribution strategy is important to mitigate the effect when certain branches cannot
Stenosis
be considered. In this study, stenotic coronary arteries from 13 patients were used to analyze the effect of missing
Fractional flow reserve
Wall shear stress
branches and different flow distribution strategies. Pressure- and wall shear stress (WSS)-derived parameters
Flow distribution around the stenotic region (ROI) were compared, including fractional flow reserve (CT-FFR), instantaneous
wave-free ratio (CT-iFR), resting distal to aortic coronary pressure (CT-Pd/Pa), time-averaged WSS, oscillatory
shear index (OSI) and relative residence time (RRT). Three flow distribution strategies were the Huo-Kassab
model at distal outlets (Type I), flow distribution based on outlet resistances (Type II), and a developed algo­
rithm distributing flow at each bifurcation until the final outlets (Type III). Results showed that Type III strategy
for models with truncated branch(es) had a good agreement in both pressure- and WSS-related results (inter­
quatile range less than 0.12% and 4.02%, respectively) with the baseline model around the ROI. The relative
difference of pressure- and WSS-related results were correlated with the flow differences in the ROI to the
baseline mode. Type III strategy had the best performance in maintaining the flow in intermediate branches. It is
recommended for CFD analysis. Removal of branches distal to a stenosis can be undertaken with an improved
performance and maintained accuracy, while those proximal to the ROI should be kept.

1. Introduction the need to consider diagnostic approaches that look beyond stenosis.
Coronary pressure-derived fractional flow reserve (FFR), which mea­
Atherosclerotic diseases are the leading cause of death and disability sures the trans-stenosis pressure difference and allows the assessment of
around the world (Libby et al., 2019). The longest established criterion blood flow impairment originating from local obstructive and down­
for risk assessment is the luminal stenosis defined by coronary computed stream perfusion, has become a powerful prognostic marker (Pijls et al.,
tomography angiography (CTA) and digital subtraction angiography 1996). The limitations of FFR measurement, especially for patients with
(DSA). However, a strong link between myocardial blood flow impair­ moderate to severe stenosis, come from its invasive nature. CTA-based
ment and adverse clinical outcomes exists beyond stenosis, highlighting and computational fluid dynamics (CFD) derived FFR (CT-FFR) has

* Corresponding authors at: Department of Radiology, Xuzhou Central Hospital, Xuzhou Clinical School of Xuzhou Medical University, No.199 Jiefang South Road,
Quanshan District, Xuzhou, Jiangsu 221009, PR China (P. Xia); Department of Radiology, University of Cambridge, Box 218, Level 5, Addenbrooke’s Hospital, Hills
Road, CB2 0QQ, Cambridge, UK (Z. Teng).
E-mail addresses: [email protected] (P. Xia), [email protected], [email protected] (Z. Teng).
1
Equal authorship.

https://doi.org/10.1016/j.jbiomech.2023.111857
Accepted 1 November 2023
Available online 3 November 2023
0021-9290/Crown Copyright © 2023 Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

consequently been developed to overcome this limitation and show a artery (LCA) were removed. In detail, the distal LAD (dLAD), the first
good correlation with invasive FFR (Nørgaard et al., 2014; Taylor et al., diagonal branch (D1) and the second diagonal branch (D2) were kept for
2013). CT-FFR can also be used to predict the effectiveness of revascu­ the LAD artery tree. The middle left circumflex artery (mLCX) and the
larisation by modeling the stented vessel and predicting the increase in first obtuse marginal branch (OM1) were kept for the LCX artery tree.
FFR (Taylor et al., 2013). In addition to the calculation of pressure, the The described LCA was used as the baseline model (model A) and its
wall shear stress (WSS) can be computed from the CTA-based CFD branch segment illustration is shown in Fig. 1c. There was a total of
analysis to improve our understanding of the development of athero­ 308,392 tetrahedral and 164,206 prismatic elements in the baseline
sclerosis by examining the interaction between the blood flow and the model A. Daughter branches of the pLAD were removed systematically
endothelium (Souilhol et al., 2020). to mimic the circumstances when there were missing branches. Models B
For a successful CFD analysis, both loading and boundary conditions to E (Fig. 1d-g) were consequently generated from model A. Model B had
at the inlet(s) and outlet(s) are needed. The aortic root is commonly used all D1, D2 and dLAD branches removed. Individual D1, dLAD and D2
as the inlet. Limited by the computational capacity and imaging reso­ were removed from model A generating models C to E, respectively.
lution, it is usually not possible to include the complete coronary Except for the changes in the LAD, all other artery trees in models B to E
vasculature so the coronary tree is selectively truncated. Moreover, the were the same as model A.
flow rate and the blood pressure are relatively easy to measure at the For the rest of the 12 models, only a single daughter branch was
aortic inlet compared with the coronary outlets (Ellwein et al., 2017). removed (Fig.S1 in Supplementary Materials). They were used to
Various technologies have been used to generate the boundary condi­ compare how much hemodynamic parameters varied between the
tions for outlets and mostly depend on local diameters, the number of missing artery models with Type III flow distribution and the complete
branches and the inlet condition (Gijsen et al., 2019). Sankaran (San­ models which are the ground truths with Type I flow distribution.
karan et al., 2016) and Sturdy (Sturdy et al., 2019) have shown that the The distal branches were removed with at least five diameters from
uncertainty due to boundary conditions (or boundary resistance) has a the removal position to the furthest ROI (defined in Section 2.4) to
large impact on the CFD simulation of common truncated models. improve the simulation efficiency. The removal of distal branches
Furthermore, the outlet boundary conditions can be explicit, such as following reference (Shi et al., 2022) demonstrates the insignificant ef­
velocity profile and outflow distribution, or implicit, such as the lumped fect of removing distal branches under certain conditions. Branches
parameter model (Gijsen et al., 2019). The average outflow distribution before and between ROIs were all kept for analysis. For the artery side
based on the inflow can be deduced by the morphological scaling laws, (left or right) without ROIs, only proximal segments were kept for
e.g., Murray’s law (Murray, 1926) or Huo-Kassab model (Huo and analysis (Shi et al., 2022).
Kassab, 2012). The lumped parameter model achieves the flow distri­
bution in artery CFD modeling similar to the current distribution in the
2.2. Computational fluid dynamic simulation
complex circuit consisting of resistance, capacitance and power supply
(Kim et al., 2010). The morphological scaling laws are still used for
The CFD simulation was performed on ADINDA 9.6 (ADINA R&D
resistance calculation, and the capacitance, mimicking the arterial
Inc, USA) and followed the procedure described in the reference (Shi
compliance, needs to be tuned for physiologically realistic coronary
et al., 2022). The blood was assumed to be Newtonian, incompressible,
waveforms. As a result, the knowledge of velocity or outflow waveforms
and viscous with density and viscosity being 1.06 g/cm3 and 0.035 P,
is necessary. Though the manual tuning of the impedances can be
respectively. A no-slip boundary condition was applied between the
difficult, there is a possible automated tuning approach (Tran et al.,
blood and the vessel wall. The flow was assumed to be transient-laminar
2017).
with the governing equation of,
There are inevitable situations in which artery models for CFD
simulation may have missing branches. These incomplete models may ∂v
ρ + ρv⋅∇v = − ∇p + μ∇2 v (1)
arise at different stages from image acquisition and segmentation to ∂t
mesh generation with or without intent. These imperfect models will where v is the flow velocity, p is the pressure, μ and ρ are blood
show a non-conservative morphological relationship with the morpho­ viscosity and density, respectively.
logical scaling laws. This study was performed to compare how pressure- The aortic inlet had time-dependent blood pressure. The aortic and
and WSS-related CFD parameters change from the complete model to coronary outlets had time-dependent blood flows. All waveforms were
the model with the missing branch(es) when three types of boundary adapted from reference (Razavi et al., 2020) and the average amount of
conditions were applied: (1) the boundary condition strategy only used the blood inflow was determined from reference (Ellwein et al., 2017).
the Huo-Kassab model for flow distribution (Type I); (2) the distributed During hyperemia, the microvascular resistance reduced to 0.24 so a
flow based on the resistance deduced from Huo-Kassab model with factor of 4.17 was applied to the blood outflow for the CT-FFR calcu­
outflow conservation (Type II); (3) the distributed flow considering each lation based on Ohm’s law (Taylor et al., 2013) and 4 % total inflow was
intermediate branch and final outlet to maintain the flow rate conser­ distributed to coronary arteries (Kim et al., 2010). The simulation was
vation (Type III). run over 5 cardiac cycles and results from the last cycle were extracted
for analysis.
2. Materials and methods

2.1. Patient-specific model reconstruction 2.3. Flow distribution strategy

A total of 13 computational models were constructed from CTA Type I flow distribution only considered the morphological rela­
images with focal stenosis with a mixture of high and low signal atten­ tionship of outlet and computed artery flow, which obeyed the Huo-
uation in the region of interest (ROI). Lumen segmentations were Kassab model (Huo and Kassab, 2012), a commonly used method
automatically generated by Aladdin CT-VascularView (Nanjing Jingsan developed from Murray’s law (Murray, 1926),
Medical Science and Technology, Ltd., China). This study was approved Q ∝1 Dδ ∝Aδ/2 (2)
by the review board of Xuzhou Central Hospital, Jiangsu, China.
A model with ~ 50 % calcified moderate stenosis at the proximal left ( )δ/2
Qbranch Abranch
anterior descending artery (pLAD) was chosen to compare the detailed = (3)
Qmain Amain
change of hemodynamic parameters based on all three types of loading
conditions (Fig. 1a). For this model, distal branches of the left coronary where Q is the blood flow rate, D is the diameter, and A is the average

2
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

(caption on next page)

3
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

Fig. 1. The chosen baseline model and the left coronary artery (LCA) models The baseline model for computational fluid dynamics analysis is shown in (a) with ROIs
for the later analysis highlighted in red outlines. A curved multiplanar reconstruction (b) of ROIs and the viewing position is indicated in (a). The baseline model A
with the LCA tree is shown in (c) with the branch segment illustration. Model B has D1, D2 and dLAD removed as shown in (d). Model C has D1 removed as shown in
(e). Model D has D2 removed as shown in (f). Model E has dLAD removed as shown in (g). Aorta and right coronary arteries were not shown in (b)-(g) to highlight the
change of coronary branches for left coronary arteries. LCA: left coronary artery; LAD: left anterior descending artery; pLAD: proximal LAD; mLAD: mid LAD; dLAD:
distal LAD; D1: the first diagonal branch; D2: the second diagonal branch; LCX: left circumflex; pLCX: proximal LCX; mLCX: mid LCX; OM1: the first obtuse marginal
branch. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

cross-sectional area measured along each coronary artery centerline (Shi 2.4. Post-processing and analysis
et al., 2022). The morphometric exponent, δ, was chosen to be 7/3 (Huo
and Kassab, 2012). Three pressure-related parameters, FFR, instantaneous wave-free
Type II flow distribution not only obeyed the Huo-Kassab model but ratio (iFR) and resting distal to aortic coronary pressure (Pd/Pa), were
also took into account the output flow conservation of the outlet distal derived from CFD analyses. First of all, the time-averaged pressure
branches. As a result, together with Eqns.2 and 3, the flow relationship (TAP) of each node was calculated over the last cardiac cycle under
of outlet for an artery tree with inflow Qin (distal branches from 1 to k), resting conditions (Lopes et al., 2020),
∫
Q1 : Q2 : ⋯ : Qk = Aδ/2 : Aδ/2 : ⋯ : Aδ/2 (4) 1 Tc
1 2 k TAP = p dt (9)
Tc 0
Qin = Σki=1 Qi (5)
where p is the pressure and Tc is the period of a cardiac cycle. CT-FFR
Type III flow distribution was developed from Type II with a top- value at a given node was calculated by dividing its mean pressure by
down strategy (Fig. 2). The conservation of flow was calculated at the mean aortic pressure during hyperemia (Koo et al., 2011),
each bifurcation rather than only considering the outflow conservation TAPnode
at the distal outlet branches. Type III flow distribution strategy is similar CT − FFR = | (10)
TAPaorta hyperemia
to the flow distribution method developed by Kishi et al. (Kishi et al.,
2018), but it was applied to multiple daughter branches and has an Apart from the FFR derived from the invasive measurement of
additional check of the morphological relationship derived from Eqns.2 pressure during hyperemia, there are other two pressure-derived
and 5 applied to the determination of flow distribution. For a parent indices, iFR and Pd/Pa, with invasive measurement. The calculation of
branch (Aparent ) without missing daughter branches (Ai ), Pd/Pa is similar to FFR but under resting conditions (Kobayashi et al.,
( )δ/2 2017). The iFR is calculated during the wave-free period under resting
Ai conditions (Sen et al., 2012). Similar to the derivation of CT-FFR from
Σki=1 ≈1 (6)
Aparent CFD simulation and FFR concepts, both iFR and Pd/Pa can be also
derived from CFD simulation (CT-iFR and CT-Pd/Pa),
( )δ/2
Ai
1 − Σki=1 < Threshold (7) TAPnode
Aparent CT − iFR = | (11)
TAPaorta restingwave− freeperiod

where the threshold depends on the accuracy of the measurement,

such as the imaging resolution. For this study, the threshold was chosen TAPnode
CT − Pd/Pa = | (12)
to be 5 %. TAPaorta resting
Instead of the direct flow distribution to all distal branches as in the Three WSS-associated parameters, time-averaged wall shear stress
Type II strategy, Type III started from the initial parent branch and kept (TAWSS), oscillatory shear index (OSI), and relative residence time
distributing flow at each bifurcation obeying the Huo-Kassab model and (RRT), were calculated for each node (Lopes et al., 2020),
flow conservation until the final outlet branches. Type III strategy, ∫
therefore, focused on keeping the flow in the intermediate artery 1 Tc
TAWSS = |τw |dt (13)
branches close to the baseline complete model. In contrast, the flow in Tc 0
an intermediate artery branch was determined by the distal daughter ⎛ ⃒∫ ⃒⎞
⃒ Tc ⃒
outlets in the Type II strategy. Bifurcations, intermediate artery 1⎝ ⃒ 0 τw dt⃒
branches and outlets were recorded when Eqn.7 was not satisfied. Those OSI = 1− ∫ Tc ⎠ (14)
2 |τw |dt
bifurcations were identified as positions of missing daughter branches.
0

All following daughter branches would have incorrect flow distribution

to the baseline complete model. RRT = [(1 − 2 × OSI) × TAWSS ]− 1
(15)
Finally, three strategies should provide the same outflows for model where τw is WSS.
A. The complete baseline model of a patient followed the intrinsic Luminal ROI i (Fig. 1b) was used to investigate the effect of flow
morphology and blood flow relationship governed by the Huo-Kassab distribution strategies on TAWSS, OSI and RRT. ROI i included the ste­
model with flow conservation at each bifurcation (Kassab, 2006) so notic luminal region and the anterior and posterior luminal regions in
the Type III strategy was satisfied at each intermediate artery until the front of and after the stenosis with the same length as the stenotic region.
outlet branches. From the scaling law of Eqn.6, there was a conservation ROI ii, 20 to 30 mm distal-to-the-lesion (Renard et al., 2019) (Fig. 1b),
of area exponentials from the first intermediate level of branches to the was chosen to investigate the effect of flow distribution strategies on CT-
final outlets and the summations at each level of arteries were all equal FFR, CT-iFR and CT-Pd/Pa. The relative difference at each node of the
to the inlet area exponential. As a result, the Type II strategy would model with missing branch(es) was assessed by referring to the value
result in the same outflows as the Type III strategy even though the Type (k = TAP, CT-FFR, CT-iFR, CT-Pd/Pa, TAWSS, OSI, RRT) of the same
II strategy only considered the outlet branches with flow conservation. node in the corresponding complete model, kA ,
At last, Type I and Type II strategies would have the same outflows since
it was intrinsic that Eqns.4 and 5 were satisfied and the distal outlet kA − kj
Δknode = × 100%|node , (j = ModelBtoE) (16)
exponentials sum was equal to the inlet area exponential as mentioned kA
above. where the subscript stands for model type (Fig. 1). Medians of rela­
tive difference for all nodes in ROI i or ii of different parameters were

4
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

Fig. 2. Type III flow distribution algorithm. A top-down strategy distributes flow at each bifurcation until the final outlets. Q is the blood flow and A is the average
area (distal branches from 1 to k).

5
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

used for comparison for different types of flow distribution strategies. 4. Discussion
Δk = median{Δknode }(node ∈ ROIiorROIii) (17)
Results showed that both subbranch(es) removal and flow distribu­
tion strategy influenced the local hemodynamic parameter calculation.
2.5. Statistical analysis Medians of relative difference for ΔCT-FFR, ΔCT-iFR and ΔCT-Pd/Pa,
were relatively minor compared to ΔTAWSS, ΔOSI and ΔRRT.
For the aforementioned chosen model comparing all three types of Invasive FFR improves the selection of patients for coronary revas­
flow distributions with systematic removal of branches, the flow dif­ cularization compared with decisions guided by stenosis alone as
ferences in ROIs between modified models and the baseline complete demonstrated by multiple trials, e.g., the DEFER (Deferral of Percuta­
model are shown in Table 1. For the other 10 models with only a single neous Coronary Intervention) (Pijls et al., 2007), the FAME (Fractional
daughter branch removed, the Wilcoxon signed-rank test was used to Flow Reserve Versus Angiography for Multivessel Evaluation) (Tonino
compare whether the distribution of median relative differences devi­ et al., 2009) and the FAME 2 (De Bruyne et al., 2012). Alternative
ated from 0 %. A p < 0.05 was considered statistically significant. invasive pressure measurements and indices are being developed with a
range of solutions such as using resting conditions, using different time
3. Results windows, and applying angiography and intravascular imaging (De
Maria et al., 2020). The use of CTA-derived CT-FFR offers a further
3.1. Comparison of three flow distribution strategies on the chosen model potential solution that avoids the need for invasive studies. A recent
meta-analysis of 5 studies including 5,460 patients examining CT-FFR
For each ROI, medians of relative differences of modified models B to derived from the truncated whole artery tree, showed that negative
E compared with model A were plotted as points to illustrate the dif­ CT-FFR results is a powerful prognostic marker of low risk for down­
ferences among each type of flow distribution (Figs. 3 and 4). The actual stream cardiovascular events (Nørgaard et al., 2022). An accurate CTA-
flow distributions at rest are provided in Fig.S2 in the Supplementary based FFR calculation is therefore important and clinically useful.
Materials. However, the analysis accuracy is greatly dependent on the corre­
Median relative differences of ΔCT-FFR, ΔCT-iFR and ΔCT-Pd/Pa, sponding flow rate which is largely determined by the branches. Un­
from models B and C with type I strategy and models D and E with Type derstanding the impact of flow rate distribution and branch(es) removal
III strategy were close to 0 % (Fig. 3). Models D and E with Type III on the CT-FFR calculation is therefore essential.
strategy varied within ~ 0.09 % for ΔCT-FFR and 0.01 % for ΔCT-iFR For the chosen single model (Fig. 1), the Type I and II flow distri­
and ΔCT-Pd/Pa. The median relative differences from the Type II bution strategy generally did not perform well in modified models with
strategy were comparatively poorer (greater than 0.5 % for ΔCT-FFR missing branches. Type III flow distribution strategy, applying
and 0.05 % for ΔCT-iFR and ΔCT-Pd/Pa) (Fig. 3). Moreover, for results morphological relationship and flow conservation for all intermediate
with the Type III strategy, models B and C both without D1 showed high branches, had all pressure- and WSS-related median relative differences
median relative differences. The box-whisker plots of relative differ­ close to 0 %, except for models B and C for pressure-related results in
ences of these three parameters are shown in Fig.S3 in Supplementary ROI ii. Table 1 shows a clue that the pressure- and WSS-related differ­
Materials. ences are correlated with the differences in blood flows of the ROIs.
There was a grouping of results among three strategies (Fig. 4 and Different strategies resulted in different flows in the ROIs and, hence
Fig.S4 in Supplementary Materials). Median relative differences of caused differences in the pressure- and WSS-related results. Type III flow
WSS-derived parameters from the Type III strategy were predominantly distribution strategy minimized the difference in the flow through the
close to 0 %. For the Type III strategy, ΔTAWSS varied from − 1.34 % to stenotic region, so its WSS-related results are the closest to baseline
0.83 %, ΔOSI from − 0.10 % to 0.95 % and ΔRRT from − 0.86 % to 1.32 results.
%. Type I strategy showed the largest deviation from 0 % with maximal Nevertheless, the removal of the D1 branch in modified models B and
ΔTAWSS, ΔOSI and ΔRRT being 38.96 %, 24.73 % and − 66.99 % from C affected pressure-related results in ROI ii (Fig. 3) with the Type III flow
model E. The box-whisker plots of relative differences of three WSS- distribution strategy. The removal of D1 caused blood flows from both
derived parameters are shown in Fig.S5 in Supplementary Materials. pLAD and D1 from the complete baseline model to flow in mLAD, as
shown in Table 1. The change in blood flow consequently altered the
3.2. Type III flow distribution strategy on models with truncated branches pressure-related results in ROI ii. In order to have a correct blood flow
distribution in ROI, there should be no branch removal before the ROI.
For the rest of the 12 models, Fig. 5 and Table 2 show the distribution The Type III algorithm (Fig. 2), however, will identify branches with the
of relative difference medians that compared results from the Type III incorrect flow automatically and the branch(es) from models with
strategy for models with a daughter branch truncated and those from the missing or removed branch(es) before the ROI will be identified and
Type I strategy for complete models. There was no statistically signifi­ warned.
cant deviation from 0 % (Table 2). The IQRs were 1.59 %, 4.02 % and Finally, the Type III flow distribution strategy is capable of dealing
1.55 %, respectively, for ΔTAWSS, ΔOSI and ΔRRT; and 0.13 %, 0.02 % with missing daughter branches with a statistically insignificant devia­
and 0.02 %, respectively, for ΔCT-FFR, ΔCT-iFR and ΔCT-Pd/Pa. tion of 0 % between the modified models and the complete models of the
other 12 cases. As a result, this novel flow distribution strategy can help
Table 1 to improve or maintain accuracy and reduce computing power and
Flow differences in ROIs i and ii with respect to the baseline model A for the complexity by removing the distal and side branches. For instance, to
chosen model.
keep the WSS-related results close to the baseline ones, the baseline
Type ROI i† ROI ii† model A with a total of 308,392 tetrahedral elements and 164,206
Model prismatic elements could be reduced (~15 %) to model B with a total of
I II III I II III
270,286 tetrahedral elements and 138,747 prismatic elements. The
B 31 % 20 % 2% 2% 19 % 45 %
C 32 % 21 % 2% 4% 17 % 45 % hyperemia simulation time was reduced from 17,017 s (about 5 h) to
D 28 % 17 % 1% 41 % 32 % 1% 13,994 s (about 4 h) with a workstation of 6 Intel(R) Core(TM) i9-10900
E 40 % 27 % 1% 59 % 50 % 1% K CPUs and 32 GB RAM.
†
ROI i included the stenotic luminal region and the anterior and posterior It is worth emphasizing that our study seeks to provide a solution to
luminal regions in front of and after the stenosis; ROI ii, 20 to 30 mm distal-to- two questions of non-invasive coronary CFD simulations in real practice:
the-lesion. (1) when and how we can remove coronary branches without affecting

6
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

Fig. 3. Medians of pressure-related relative differences in ROI ii of the chosen model. FFR: fractional flow reserve; iFR: instantaneous wave-free ratio; Pd/Pa: resting
distal to aortic coronary pressure.

Fig. 4. Medians of wall shear stress(WSS)-related relative differences in ROI i of the chosen mode. TAWSS: time-averaged wall shear stress; OSI: oscillatory shear
index; RRT: relative residence time.

Fig. 5. Box-whisker plots for wall shear stress (WSS)-and pressure-related relative differences of 12 models. FFR: fractional flow reserve; iFR: instantaneous wave-
free ratio; OSI: oscillatory shear index; Pd/Pa: resting distal to aortic coronary pressure; RRT: relative residence time TAP: time-averaged pressure; TAWSS: time-
averaged wall shear stress.

the WSS and pressure in the ROIs; (2) whether the segmentation model than 5 diameters away from the position of removal (Shi et al., 2022).
or the image resolution is good enough for the CFD simulation. For the For the second question, there will be more confidence and less time
first question, this study suggests caution should be used in removing spent on the ambiguous cases with the Type III strategy that checks the
branches proximal to the ROI, although it is possible to remove small flow conservation at the ROI based on the provided segmentations.
proximal branches when a threshold is chosen. The Type III strategy will There will be considerable variation in the ROI, at least in WSS, from the
show warnings if the threshold is reached. There is more flexibility in the CFD simulation once the flow distribution at the ROI has reached the
removal of distal branches with the Type III strategy if the ROI is more pre-defined threshold.

7
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

Table 2 for CFD loading condition calculations. Removal of branches distal to a

Relative difference medians of 12 models comparing the missing artery models stenosis can be undertaken with improved performance and maintained
with Type III flow distribution and the complete models with Type I flow accuracy, while removal of branches proximal to the ROI is not
distribution. recommended.
Parameter Median [Range] P Value

ΔTAWSS − 0.13 % [ − 2.49 %, 1.72 %] 0.54 CRediT authorship contribution statement

ΔOSI 0.19 % [ − 2.28 %, 3.68 %] 0.84
ΔRRT − 0.13 % [ − 1.67 %, 2.43 %] 0.89 Yibing Shi: Writing – review & editing, Resources, Formal analysis,
0.05 % [ − 1.07 %, 0.58 %] 0.50
Data curation. Jin Zheng: Writing – review & editing, Writing – original
ΔFFR
ΔiFR − 0.00 % [ − 0.13 %, 0.11 %] 0.67
ΔPd/Pa − 0.00 % [ − 0.10 %, 0.10 %] 0.88 draft, Visualization, Methodology, Formal analysis. Ying Zhang:
Writing – review & editing, Formal analysis, Data curation. Quanlin
FFR: fractional flow reserve; iFR: instantaneous wave-free ratio; OSI: oscillatory
Sun: Visualization, Formal analysis. Jinhua Shen: Writing – review &
shear index; Pd/Pa: resting distal to aortic coronary pressure; RRT: relative
editing, Software, Resources, Investigation, Formal analysis. Yong­
residence time TAP: time-averaged pressure; TAWSS: time-averaged wall shear
stress. guang Gao: Visualization, Software, Data curation. Jingxi Sun: Writing
– review & editing, Validation, Software, Formal analysis, Data curation.
Ning Yang: Writing – review & editing, Formal analysis, Data curation.
Limitations exist in this study. Firstly, there are uncertainties related
Xuanxuan Zhou: Writing – review & editing, Formal analysis. Suqing
to the choice of morphometric exponent, δ (Eqs. 2–4, 6, 7). van der
Li: Writing – review & editing, Software, Formal analysis, Data curation.
Giessen (van der Giessen et al., 2011) has studied the dependency of
Jonathan R. Weir-McCall: Writing – review & editing, Supervision.
WSS calculation on δ (3 and 2.55). Further study needs to investigate the
Ping Xia: Writing – review & editing, Formal analysis, Data curation.
robustness of the current algorithm to the δ value. Additionally, there is
Zhongzhao Teng: Writing – review & editing, Supervision, Project
less difference between 7/3 in our study and 2.55 (8 %) from the
administration, Methodology, Funding acquisition, Conceptualization.
empirical model (van der Giessen et al., 2011) compared with 3 (18 %)
in Murray’s law (Murray, 1926). A following study by Lodi Rizzini (Lodi
Rizzini et al., 2022) has shown the difference between the δ value and Declaration of Competing Interest
the invasive measurements. While there are promising ways to measure
blood flow invasively, this study focuses more on the non-invasive The authors declare that they have no known competing financial
application of the coronary CFD simulations with the non-invasive interests or personal relationships that could have appeared to influence
measurement or quantification of blood flow still a challenge. Sec­ the work reported in this paper.
ondly, 4 % flow distribution to coronary arteries may not hold for the
resting condition. Further study should be carried out to validate this Acknowledgments
assumption and may implement the personalized flow calculation (e.g.,
the average total coronary inflow) based on the myocardial mass from This research was supported by the Bureau of Science and Technol­
the allometric scaling law (Choy and Kassab, 2008). Thirdly, in our ogy, Xuzhou, China (KC19176), and the NIHR Cambridge Biomedical
chosen model, we observed different effects on blood pressure Research Centre (BRC-1215-20014).
(maximum around 2 % in ΔCT-FFR) and the WSS (maximum around 65
% in ΔRRT) from the same change of the flow ratio. Further study should
Appendix A. Supplementary data
be carried out to investigate how the change of flow in the stenosis is
related to the change in WSS and distal blood pressure. In addition, for a
Supplementary data to this article can be found online at https://doi.
fully developed, incompressible and laminar pipe flow, the local WSS is
org/10.1016/j.jbiomech.2023.111857.
correlated with the radial flow velocity gradient and the change of
pressure is correlated with the change of the square of the longitudinal
References
flow velocity from the Darcy–Weisbach equation. Besides, there is a non-
linear relationship between the CT-FFR and the level of stenosis (San­ Choy, J.S., Kassab, G.S., 2008. Scaling of myocardial mass to flow and morphometry of
karan et al., 2016). The level of stenosis also challenges the modeling coronary arteries. J. Appl. Physiol. 104, 1281–1286. https://doi.org/10.1152/
assumption of transient-laminar flow (Freidoonimehr et al., 2020; japplphysiol.01261.2007.
De Bruyne, B., Pijls, N.H.J., Kalesan, B., Barbato, E., Tonino, P.A.L., Piroth, Z., Jagic, N.,
Mahalingam et al., 2016). Fourthly, the Type III strategy is expected to
Möbius-Winkler, S., Rioufol, G., Witt, N., Kala, P., MacCarthy, P., Engström, T.,
work better when distal branches are missing but proximal branches are Oldroyd, K.G., Mavromatis, K., Manoharan, G., Verlee, P., Frobert, O., Curzen, N.,
preserved. Further study should be carried out to investigate the per­ Johnson, J.B., Jüni, P., Fearon, W.F., 2012. Fractional Flow Reserve-Guided PCI
versus Medical Therapy in Stable Coronary Disease. N. Engl. J. Med. 367, 991–1001.
formance of the Type III strategy in some extreme cases, e.g. very diffuse
https://doi.org/10.1056/NEJMoa1205361.
disease in multiple vascular segments. The segmentation accuracy De Maria, G.L., Garcia-Garcia, H.M., Scarsini, R., Hideo-Kajita, A., Gonzalo López, N.,
should be the main challenge. The Type III flow distribution may be used Leone, A.M., Sarno, G., Daemen, J., Shlofmitz, E., Jeremias, A., Tebaldi, M., Bezerra,
to check the segmentation as the flow conservation should still hold. The H.G., Tu, S., Lemos, P.A., Ozaki, Y., Dan, K., Collet, C., Banning, A.P., Barbato, E.,
Johnson, N.P., Waksman, R., 2020. Novel Indices of Coronary Physiology: Do We
challenge to both segmentation and flow conservation is when there is a Need Alternatives to Fractional Flow Reserve? Circ. Cardiovasc. Interv. https://doi.
new development of compensatory small vessels that are comparable to org/10.1161/CIRCINTERVENTIONS.119.008487.
CTA resolution. Fifthly, further study should compare the flow distri­ Ellwein, L., Samyn, M.M., Danduran, M., Schindler-Ivens, S., Liebham, S., LaDisa, J.F.,
2017. Toward translating near-infrared spectroscopy oxygen saturation data for the
bution based on transluminal attenuation gradient (TAG) (Kishi et al., non-invasive prediction of spatial and temporal hemodynamics during exercise.
2018; Müller et al., 2021), but the proportionality between flow and Biomech. Model. Mechanobiol. 16, 75–96. https://doi.org/10.1007/s10237-016-
TAG might not hold for missing or removing branch situations. More­ 0803-4.
Freidoonimehr, N., Arjomandi, M., Sedaghatizadeh, N., Chin, R., Zander, A., 2020.
over, CFD analysis was performed with the assumption of a rigid wall. Transitional turbulent flow in a stenosed coronary artery with a physiological
Wall deformation due to dynamic blood pressure and myocardium pulsatile flow. Int. J. Numer. Method. Biomed. Eng. 36, 1–25. https://doi.org/
contraction were not considered. If the CT-FFR is the only measurement 10.1002/cnm.3347.
Gijsen, F., Katagiri, Y., Barlis, P., Bourantas, C., Collet, C., Coskun, U., Daemen, J.,
of interest, the reduced-order 1D model of leakage function from Gosling
Dijkstra, J., Edelman, E., Evans, P., Van Der Heiden, K., Hose, R., Koo, B.K.,
et al. (Gosling et al., 2020) also showed good performance in 146 ar­ Krams, R., Marsden, A., Migliavacca, F., Onuma, Y., Ooi, A., Poon, E., Samady, H.,
teries from 80 patients with incomplete models. Stone, P., Takahashi, K., Tang, D., Thondapu, V., Tenekecioglu, E., Timmins, L.,
Torii, R., Wentzel, J., Serruys, P., 2019. Expert recommendations on the assessment
In conclusion, the Type III flow distribution strategy is recommended
of wall shear stress in human coronary arteries: Existing methodologies, technical

8
Y. Shi et al. Journal of Biomechanics 161 (2023) 111857

considerations, and clinical applications. Eur. Heart J. 40, 3421–3433. https://doi. 2022. Prognostic value of coronary computed tomography angiographic derived
org/10.1093/eurheartj/ehz551. fractional flow reserve: a systematic review and meta-analysis. Heart 108, 194–202.
Gosling, R.C., Sturdy, J., Morris, P.D., Fossan, F.E., Hellevik, L.R., Lawford, P., Hose, D. https://doi.org/10.1136/heartjnl-2021-319773.
R., Gunn, J., 2020. Effect of side branch flow upon physiological indices in coronary Pijls, N.H.J., de Bruyne, B., Peels, K., van der Voort, P.H., Bonnier, H.J.R.M.,
artery disease. J. Biomech. 103, 109698 https://doi.org/10.1016/j. Bartunek, J., Koolen, J.J., 1996. Measurement of Fractional Flow Reserve to Assess
jbiomech.2020.109698. the Functional Severity of Coronary-Artery Stenoses. N. Engl. J. Med. 334,
Huo, Y., Kassab, G.S., 2012. Intraspecific scaling laws of vascular trees. J. r. Soc. Interface 1703–1708. https://doi.org/10.1056/nejm199606273342604.
9, 190–200. https://doi.org/10.1098/rsif.2011.0270. Pijls, N.H.J., van Schaardenburgh, P., Manoharan, G., Boersma, E., Bech, J.-W., van’t
Kassab, G.S., 2006. Scaling laws of vascular trees: of form and function. Am. J. Physiol. Veer, M., Bär, F., Hoorntje, J., Koolen, J., Wijns, W., de Bruyne, B.,, 2007.
Circ. Physiol. 290, H894–H903. https://doi.org/10.1152/ajpheart.00579.2005. Percutaneous Coronary Intervention of Functionally Nonsignificant Stenosis. J. Am.
Kim, H.J., Vignon-Clementel, I.E., Coogan, J.S., Figueroa, C.A., Jansen, K.E., Taylor, C.A., Coll. Cardiol. 49, 2105–2111. https://doi.org/10.1016/j.jacc.2007.01.087.
2010. Patient-specific modeling of blood flow and pressure in human coronary Razavi, A., Sachdeva, S., Frommelt, P.C., LaDisa, J.F., 2020. Patient-Specific Numerical
arteries. Ann. Biomed. Eng. 38, 3195–3209. https://doi.org/10.1007/s10439-010- Analysis of Coronary Flow in Children with Intramural Anomalous Aortic Origin of
0083-6. Coronary Arteries. Semin. Thorac. Cardiovasc. Surg. 1–13. https://doi.org/10.1053/
Kishi, S., Giannopoulos, A.A., Tang, A., Kato, N., Chatzizisis, Y.S., Dennie, C., j.semtcvs.2020.08.016.
Horiuchi, Y., Tanabe, K., Lima, J.A.C., Rybicki, F.J., Mitsouras, D., 2018. Fractional Renard, B.M., Cami, E., Jiddou-Patros, M.R., Said, A., Kado, H., Trivax, J., Berman, A.,
flow reserve estimated at coronary CT angiography in intermediate lesions: Gulati, A., Rabah, M., Timmis, S., Shoukfeh, M., Abbas, A.E., Hanzel, G., Hanson, I.,
Comparison of diagnostic accuracy of different methods to determine coronary flow Dixon, S., Safian, R.D., 2019. Optimizing the Technique for Invasive Fractional Flow
distribution. Radiology 287, 76–84. https://doi.org/10.1148/radiol.2017162620. Reserve to Assess Lesion-Specific Ischemia. Circ. Cardiovasc. Interv. 12, 1–9. https://
Kobayashi, Y., Johnson, N.P., Zimmermann, F.M., Witt, N., Berry, C., Jeremias, A., doi.org/10.1161/CIRCINTERVENTIONS.119.007939.
Koo, B.K., Esposito, G., Rioufol, G., Park, S.J., Nishi, T., Choi, D.H., Oldroyd, K.G., Sankaran, S., Kim, H.J., Choi, G., Taylor, C.A., 2016. Uncertainty quantification in
Barbato, E., Pijls, N.H.J., De Bruyne, B., Fearon, W.F., 2017. Agreement of the coronary blood flow simulations: Impact of geometry, boundary conditions and
Resting Distal to Aortic Coronary Pressure With the Instantaneous Wave-Free Ratio. blood viscosity. J. Biomech. 49, 2540–2547. https://doi.org/10.1016/j.
J. Am. Coll. Cardiol. 70, 2105–2113. https://doi.org/10.1016/j.jacc.2017.08.049. jbiomech.2016.01.002.
Koo, B.K., Erglis, A., Doh, J.H., Daniels, D.V., Jegere, S., Kim, H.S., Dunning, A., Sen, S., Escaned, J., Malik, I.S., Mikhail, G.W., Foale, R.A., Mila, R., Tarkin, J.,
Defrance, T., Lansky, A., Leipsic, J., Min, J.K., 2011. Diagnosis of ischemia-causing Petraco, R., Broyd, C., Jabbour, R., Sethi, A., Baker, C.S., Bellamy, M., Al-
coronary stenoses by noninvasive fractional flow reserve computed from coronary Bustami, M., Hackett, D., Khan, M., Lefroy, D., Parker, K.H., Hughes, A.D.,
computed tomographic angiograms: Results from the prospective multicenter Francis, D.P., Di Mario, C., Mayet, J., Davies, J.E., 2012. Development and validation
DISCOVER-FLOW (Diagnosis of Ischemia-Causing Stenoses Obtained Via Noni. of a new adenosine-independent index of stenosis severity from coronary wave-
J. Am. Coll. Cardiol. 58, 1989–1997. https://doi.org/10.1016/j.jacc.2011.06.066. intensity analysis: Results of the ADVISE (ADenosine Vasodilator Independent
Libby, P., Buring, J.E., Badimon, L., Hansson, G.K., Deanfield, J., Bittencourt, M.S., Stenosis Evaluation) study. J. Am. Coll. Cardiol. 59, 1392–1402. https://doi.org/
Tokgözoğlu, L., Lewis, E.F., 2019. Atherosclerosis. Nat. Rev. Dis. Prim. 5, 1–18. 10.1016/j.jacc.2011.11.003.
https://doi.org/10.1038/s41572-019-0106-z. Shi, Y., Zheng, J., Yang, N., Chen, Y., Sun, J., Zhang, Y., Zhou, X., Gao, Y., Li, S., Zhu, H.,
Lodi Rizzini, M., Candreva, A., Chiastra, C., Gallinoro, E., Calò, K., D’Ascenzo, F., De Acosta-Cabronero, J., Xia, P., Teng, Z., 2022. The effect of subbranch for the
Bruyne, B., Mizukami, T., Collet, C., Gallo, D., Morbiducci, U., 2022. Modelling quantification of local hemodynamic environment in the coronary artery: A
coronary flows: impact of differently measured inflow boundary conditions on computed tomography angiography-based computational fluid dynamic analysis.
vessel-specific computational hemodynamic profiles. Comput. Methods Programs Emerg. Crit. Care Med. 2, 181–190. https://doi.org/10.1097/
Biomed. 221, 106882 https://doi.org/10.1016/j.cmpb.2022.106882. EC9.0000000000000062.
Lopes, D., Puga, H., Teixeira, J., Lima, R., 2020. Blood flow simulations in patient- Souilhol, C., Serbanovic-Canic, J., Fragiadaki, M., Chico, T.J., Ridger, V., Roddie, H.,
specific geometries of the carotid artery: A systematic review. J. Biomech. 111, Evans, P.C., 2020. Endothelial responses to shear stress in atherosclerosis: a novel
110019 https://doi.org/10.1016/j.jbiomech.2020.110019. role for developmental genes. Nat. Rev. Cardiol. 17, 52–63. https://doi.org/
Mahalingam, A., Gawandalkar, U.U., Kini, G., Buradi, A., Araki, T., Ikeda, N., 10.1038/s41569-019-0239-5.
Nicolaides, A., Laird, J.R., Saba, L., Suri, J.S., 2016. Numerical analysis of the effect Sturdy, J., Kjernlie, J.K., Nydal, H.M., Eck, V.G., Hellevik, L.R., 2019. Uncertainty
of turbulence transition on the hemodynamic parameters in human coronary quantification of computational coronary stenosis assessment and model based
arteries. Cardiovasc. Diagn. Ther. 6, 208–220. https://doi.org/10.21037/ mitigation of image resolution limitations. J. Comput. Sci. 31, 137–150. https://doi.
cdt.2016.03.08. org/10.1016/j.jocs.2019.01.004.
Müller, L.O., Fossan, F.E., Bråten, A.T., Jørgensen, A., Wiseth, R., Hellevik, L.R., 2021. Taylor, C.A., Fonte, T.A., Min, J.K., 2013. Computational fluid dynamics applied to
Impact of baseline coronary flow and its distribution on fractional flow reserve cardiac computed tomography for noninvasive quantification of fractional flow
prediction. Int. J. Numer. Method. Biomed. Eng. 37 https://doi.org/10.1002/ reserve: Scientific basis. J. Am. Coll. Cardiol. 61, 2233–2241. https://doi.org/
cnm.3246. 10.1016/j.jacc.2012.11.083.
Murray, C.D., 1926. The Physiological Principle of Minimum Work: I. The Vascular Tonino, P.A.L., De Bruyne, B., Pijls, N.H.J., Siebert, U., Ikeno, F., van ‘t Veer, M., Klauss,
System and the Cost of Blood Volume. Proc. Natl. Acad. Sci. 12, 207–214. https:// V., Manoharan, G., Engstrøm, T., Oldroyd, K.G., Ver Lee, P.N., MacCarthy, P.A.,
doi.org/10.1073/pnas.12.3.207. Fearon, W.F.,, 2009. Fractional Flow Reserve versus Angiography for Guiding
Nørgaard, B.L., Leipsic, J., Gaur, S., Seneviratne, S., Ko, B.S., Ito, H., Jensen, J.M., Percutaneous Coronary Intervention. N. Engl. J. Med. 360, 213–224. https://doi.
Mauri, L., De Bruyne, B., Bezerra, H., Osawa, K., Marwan, M., Naber, C., Erglis, A., org/10.1056/NEJMoa0807611.
Park, S.J., Christiansen, E.H., Kaltoft, A., Lassen, J.F., Bøtker, H.E., Achenbach, S., Tran, J.S., Schiavazzi, D.E., Ramachandra, A.B., Kahn, A.M., Marsden, A.L., 2017.
2014. Diagnostic performance of noninvasive fractional flow reserve derived from Automated tuning for parameter identification and uncertainty quantification in
coronary computed tomography angiography in suspected coronary artery disease: multi-scale coronary simulations. Comput. Fluids 142, 128–138. https://doi.org/
The NXT trial (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps). 10.1016/j.compfluid.2016.05.015.
J. Am. Coll. Cardiol. 63, 1145–1155. https://doi.org/10.1016/j.jacc.2013.11.043. van der Giessen, A.G., Groen, H.C., Doriot, P.A., de Feyter, P.J., van der Steen, A.F.W.,
Nørgaard, B.L., Gaur, S., Fairbairn, T.A., Douglas, P.S., Jensen, J.M., Patel, M.R., van de Vosse, F.N., Wentzel, J.J., Gijsen, F.J.H., 2011. The influence of boundary
Ihdayhid, A.R., Ko, B.S.H., Sellers, S.L., Weir-McCall, J., Matsuo, H., Sand, N.P.R., conditions on wall shear stress distribution in patients specific coronary trees.
Øvrehus, K.A., Rogers, C., Mullen, S., Nieman, K., Parner, E., Leipsic, J., Abdulla, J., J. Biomech. 44, 1089–1095. https://doi.org/10.1016/j.jbiomech.2011.01.036.