Medical Engineering and Physics 78 (2020) 1–13

Contents lists available at ScienceDirect

Medical Engineering and Physics

journal homepage: www.elsevier.com/locate/medengphy

Fluid–structure interaction modeling in cardiovascular medicine – A

systematic review 2017–2019
Matthew Hirschhorn a, Vakhtang Tchantchaleishvili b, Randy Stevens c,d, Joseph Rossano e,
Amy Throckmorton a,∗
a
BioCirc Research Laboratory, School of Biomedical Engineering, Science, and Health Systems, Drexel University, Philadelphia, PA USA
b
Division of Cardiothoracic Surgery, Thomas Jefferson University, Philadelphia, PA USA
c
St. Christopher’s Hospital for Children, Tower Health Hospitals, Philadelphia, PA USA
d
Pediatrics, Drexel University College of Medicine, Philadelphia, PA USA
e
Division of Pediatric Cardiology, The Children’s Hospital of Philadelphia, Philadelphia, PA, USA

a r t i c l e i n f o a b s t r a c t

Article history: Finite element analysis (FEA) and computational fluid dynamics (CFD) are generally insufficient indepen-
Received 11 August 2019 dently to model the physics of the cardiovascular system. Individually, they are unable to resolve the
Revised 18 January 2020
interplay between the solid and fluid domains, and the interplay is integral to the functioning of the sys-
Accepted 26 January 2020
tem. The use of fluid–structure interaction (FSI) methods overcomes these shortcomings by providing the
means to couple the fluid and structural domains. In the last decade, the utilization of FSI has greatly
Keywords: increased in cardiovascular engineering. In this study, we conducted a systematic review process of more
Fluid structure interaction than 10 0 0 journal articles to investigate the implementation of One-Way and Two-Way FSI for cardio-
Cardiovascular modeling vascular applications. We explored the utility of FSI to study aneurysms, the hemodynamics of patient
FSI
anatomies, native and prosthetic heart valve dynamics, flow and hemodynamics of blood pumps, and
Computational modeling
atherosclerosis. Computational resource requirements, implementation strategies and future directions of
CFD
Finite element analysis FSI for cardiovascular applications are also discussed.
Cardiovascular mechanics © 2020 Published by Elsevier Ltd on behalf of IPEM.
FEA
Numerical simulations

1. Introduction [2]. FSI has emerged as a valuable tool in cardiovascular engineer-

ing over the past 20 years, and this systematic review examines
Fluid–structure interaction (FSI) studies combine computational the application of FSI to cardiovascular engineering and medicine
fluid dynamics (CFD) with finite element analysis (FEA), enabling (Fig. 1).
the investigation of fluid behavior, structural behavior, and how FSI strategies can be implemented on geometries of interest
they interact and affect each other. Traditional CFD, using a finite and allow for the creation of patient specific models from MRI
volume approach, has limitations in exploring challenges in car- or CT imaging data. In general, current cardiovascular FSI stud-
diovascular engineering because it cannot represent how the fluid ies are completed using one or more patient specific geometries.
will respond to an adjacent structure deforming due to fluid pres- The slices from 2D image scans are stacked and merged, creat-
sure or other forces [1]. Vessel elasticity and compliance are inte- ing a 3D patient specific geometry [1,3,4]. These methods are well
gral characteristics of the circulatory system and are not modeled established in cardiovascular CFD and are being leveraged to cre-
well with CFD. Likewise, FEA methods cannot model the struc- ate more physiologically accurate FSI models. FSI also allows for
tural response to fluid dynamics if the physics of the fluid are not the study of physical phenomena that are not easily measured
known and are inherently dependent on the behavior of the struc- and are relevant factors in disease initiation, progression, diagnosis,
ture itself. FSI leverages established finite volume-based CFD and or treatment. For example, the direct measurement of wall shear
structural FEA computational methods and couples the physics of stress (WSS) in a blood vessel within the body is impossible, yet
the fluid and structural domains through the CFD and FEA solvers pathologically low WSS is a major risk factor for initiation and pro-
gression of atherosclerosis [5,6]. FSI enables clinical teams to con-
sider the impact of parameters, such as WSS, and to make treat-
∗
Corresponding author. ment decisions based on patient specific models.
E-mail address: [email protected] (A. Throckmorton).

https://doi.org/10.1016/j.medengphy.2020.01.008
1350-4533/© 2020 Published by Elsevier Ltd on behalf of IPEM.
2 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

Fig. 1. Examples of Fluid-Structure Interaction Studies in Cardiovascular Medicine, including studies focused on disease specific modeling [113], patient specific modeling
[36], stents [112], heart pump dynamics [124] and valve dynamics [70].

There is significant published literature focused on the develop-

ment and validation of the mathematical strategies that underlie
FSI. These studies aim to either improve model accuracy and
validity and/or decrease the required computational time and
power. Our investigation focuses on how FSI is being employed to
model and study applications in cardiovascular engineering and
medicine. This review begins with background on the systemic
review methods, general FSI implementation and the required
computational resources for different types of FSI models. We then
explore the utility of FSI to study aneurysms, the hemodynamics
of patient anatomies, native and prosthetic heart valve dynamics,
flow and hemodynamics of blood pumps, and atherosclerosis.

2. Systematic review methods

We utilized a systematic review process whereby multiple

combinations of keywords were entered into electronic databases
and used to locate relevant articles. Databases searched included Fig. 2. Process of systematic literature review. (a) Overview of methodology (b)
GoogleScholar, PubMed, Web of Science, and ProQuest Research Li- quantification of results of systemic literature review.
brary. Combinations of keywords consisted of FSI, fluid–structure
interaction, fluid structure interaction, cardio, cardiac, cardiovascu-
lar, vascular, stent, mechanical circulatory assist, heart pump, ven- diac pathology, or a cardiac device, (2) was not exclusively focused
tricular assist device, VAD, heart valve, patient specific, atheroscle- on the development of fluid–structure interaction methods, (3) had
rosis, and aneurysm. Our search was constrained to articles that clinical relevance, and (4) was published in English in a peer-
are electronically available and published since 2017, representing reviewed journal. Database search results were initially screened
the cutting edge application of FSI. by reviewing titles and abstracts, and the full-text articles were
Articles for inclusion met the following criteria: (1) used FSI to then reviewed for relevance and screened for quality. Fig. 2 il-
investigate a patient specific cardiac pathology, an idealized car- lustrates the selection process and methodology used in this re-
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 3

view. The decision to focus on studies published since 2017 pre- solved and require more computational power and the use of cou-
cludes the inclusion of much of the pioneering FSI work completed pling algorithms to link the two domains. Both One-Way and Two-
by Drs. Thomas J.R. Hughes, Tayfun Tezduyar, and Yuri Bazilevs, Way FSI studies have strong utility and the method selected de-
among others [7–11]. A selection of their work is referenced to pends on the problem being modeled and the resources available
honor their many and significant contributions. to the researcher.

3. General implementation strategy 4. One-Way coupled FSI studies

FSI simulations are rooted in complex multiphysical phenom- In a One-Way study, either the fluid or structural domain is
ena, thus requiring advanced discretization methods for solvers. initially solved independently, and the results are employed as a
Two primary solver-based approaches are used: monolithic and model condition when solving the other domain. An example of a
partitioned. In the monolithic approach, the fluid and structural cardiovascular model appropriate for a One-Way FSI study is the
domains are meshed together, and a single set of governing equa- modeling vessel compliance and distention, whereby the fluid do-
tions are applied to both domains concurrently. This results in bet- main is first solved using CFD to determine the fluid forces at ves-
ter accuracy at the expense of more computational investment. The sel wall and then the results are used to calculate the wall defor-
most notable monolithic approach used in cardiovascular engineer- mation using FEA. An interpolation step will couple the two do-
ing is the immersed-boundary (IB) method. Monolithic approaches, mains and remap the fluid forces from the nodes of the fluid do-
such as IB, can also be classified as non-conforming mesh methods main to the nodes of the structural domain. One-Way studies have
because the physics at boundary between the fluid and structural the advantage of model simplicity and require fewer computational
domains is represented as constraints in the model equations. IB resources, but they do not fully capture the interplay of the fluid
methods have seen an increase in utility as methods have become and structural domains [17,19].
refined and computing power more readily available [12–14]. They In the previous example, in a true in vivo case, the distention
are popular for heart valve simulations because of the thin valve of the vessel will immediately reduce the fluid pressure experi-
leaflets experience high levels of structural deformation [15,16]. enced by the vessel, thus the fluid pressure exported to the struc-
This is well suited to IB where remeshing is not required, and the tural domain are likely artificially high, resulting in an overestima-
structural mesh is within the larger model mesh. tion of distention. A major limitation of One-Way studies is that
In contrast to the monolithic approach, partitioned approaches they cannot study transient problems because the state of one do-
treat the fluid and structural domains as independent, each with main at a static point in time is used to determine the state of
their own mesh and set of govern equations. Partitioned methods the other domain at the same moment [17]. Another limitation of
require an interface between the two domains to facilitate data One-Way FSI studies is that the structural mesh does not change
transfer. These usually require remeshing as the geometries of the as the body deforms, and during significant structural deformation
fluid and solid domains change shape [13]. A major advantage this can result in highly skewed mesh elements that create un-
of partitioned methods is that they can leverage advanced CFD stable simulations that are difficult to converge and are prone to
and/or FEA techniques that have been developed independent errors. The cardiovascular system has significant transient and de-
of each other. This allows complex problems to be studied with formative behavior, especially in the heart and vasculature; thus,
partitioned methods. Calculation of the location of the interface two-way studies are better suited if a transient solution has clinical
and the transfer of information between domains requires ad- relevance. Fig. 3(a) shows an overview of One-Way FSI information
ditional computational resources and can introduce errors. For flow.
this reason, significant effort has focused on automatic remesh-
ing, interface accuracy and efficiency [13]. These are known as
5. Two-Way FSI
conforming mesh methods because new meshes are created as
the simulation steps through time, so that the boundary be-
In a Two-Way FSI study, the fluid and structural domains are
tween the fluid and structural domains can be updated as the
solved in parallel. At each sub-step the fluid and structural domain
geometries change. At each time step, an iterative approach is
solutions are required to converge before moving to the next step.
used to elucidate convergence between the structural and fluid
The fluid and structural domains are linked through a coupling
domains. A popular technique that uses a partitioned, conform-
scheme, like the ALE method, where the mesh of one domain is al-
ing mesh is the Arbitrary–Lagrangian–Eulerian (ALE) method
lowed deform to adhere to the boundaries of the other domain, or
[13,17,18].
IB method, where the mesh does not deform and structural move-
ment within the fluid is accounted for by adding body forces to the
3.1. FSI domain coupling strategies equations of motion [13,14,21]. Fully-coupled Two-Way FSI stud-
ies are typically transient and capture fluid/structure behavior over
The FSI solutions are driven based on the partitioned method a period. An example of a system requiring Two-Way FSI is the
used; the next consideration is how data are exchanged at the study of heart valve dynamics. In the native heart, valve leaflet dy-
boundary or interface between the fluid solver and structural namics are dependent upon the pressure states in the heart and
solver. This fluid–structure interface requires a coupling mecha- the attached vessels and the pressure states and fluid flow are
nism to appropriately time and feed information to the respec- dependent on the dynamic state of the valves and heart muscle
tive solvers. There are two strategies for FSI coupling between the [15,16,22]. A One-Way FSI study would be unable to capture the
structural and fluid domain physics: One-Way and fully-coupled fluid physics, heart muscle and heart valves over time. The num-
Two-Way. In a One-Way FSI study, the fluid and structural domains ber of Two-Way FSI studies has increased recently and many of
are solved in series, where the solution of one domain is used as papers reviewed in this article utilize a Two-Way technique. These
a boundary or initial condition in the second domain [17,19]. In studies are often completed on high performance computing (HPC)
a fully-coupled two-way FSI study, the two domains are solved in clusters. The additional computational demands of using Two-Way
parallel, and the solution at each iteration for the fluid and struc- FSI is justified when One-Way is not accurate enough and/or must
tural domains must agree and reach convergence together [17,20]. have a transient component. Fig. 3(b) shows an overview of Two-
Fully-coupled Two-Way studies enable transient problems to be Way FSI information flow.
4 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

Fig. 3. Explanation of (a) One-Way and (b) Two-Way FSI methodology.

Fig. 4. Mesh generation and quality workflow (a)–(c) example aspect ratio, Jacobian ratio and skewness for 2D quadrilaterals and triangles. (d) example warping factor (e)
example mess independence study.

6. Mesh generation and quality finement will yield essentially the same solution, and the solution
can be considered independent of mesh density. In a grid indepen-
One of the first steps when completing CFD, FEA or FSI stud- dence study, solution components of high interest are monitored as
ies is mesh generation, where the fluid and structural domains are the mesh is sequentially refined until the results are unchanging.
divided into 3D grids of discrete elements, the nodes of which As a rule, finer meshes require more computational resources, thus
are where the governing equations are applied and numerically it is important to determine the mesh required for independence
solved using CFD or FEA techniques. This means mesh quality is of to minimize simulation runtime. Fig. 4(e) illustrates an example of
paramount importance. Mesh quality metrics such as Jacobian ra- a mesh independence study results.
tio, skewness and aspect ratio are used to ascertain the quality of The final consideration when generating meshes for partitioned
the mesh [23]. Jacobian ratio, most important to structural simula- FSI models is an attempt to match the mesh density at the inter-
tions, is a measure of element distortion and should be minimized face between the fluid and structural domains because the surface
to increase the accuracy of mapping a real space to numerical val- nodes will not necessarily align, requiring an interpolation step to
ues. Aspect ratio is a measure of element stretch and is calculated transfer state conditions between the two domains. Interpolation
as the ratio of the maximum to minimum surface area of elements. errors will be minimized if the mesh density is similar between
Finally, skewness is a measure of how close an element is to hav- the two domains.
ing equilateral or equiangular sides, with a low value indicating a
more equilateral element. Elements with high skewness values are 7. Computational resource requirements
almost coplanar slivers, and their inclusion will introduce errors
[23]. These metrics should be assessed for each mesh, and areas Limitations in computational resources are a constraint on
that are of poor quality should be refined. Fig. 4(a)–(d) provides a study complexity and variability of model conditions. The time
visual representation of the different mesh metrics [23]. it takes to complete FSI studies is dependent on several factors:
Another important step to ensure a high-quality mesh is to per- model complexity, available computing power, and the compu-
form a grid independence study [24–26]. A geometry can be di- tational methods leveraged. Chiu et al. [27] used an FSI study
vided into elements of any size and the use of smaller elements to investigate the thrombogenicity of two commercially available
will increase accuracy up to a point. Beyond that point further re- ventricular assist devices (VADs). The model had significant com-
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 5

plexity that included modeling the flow of blood and suspended vessel and aneurysm geometry on rupture risk. The study inves-
platelets through VADs. The study leveraged a powerful parallel tigated the role of both DAAA and biomechanical parameters on
computing setup that could utilize up to 30 cores. Despite the rupture risk, confirming that DAAA is insufficient without including
available power, each simulation took an incredible amount of markers such as WSS and principal stress when making treatment
time to complete, approximately 4 months [27]. In another pub- decisions. In another model investigating the role of AA geome-
lished study by Chen et al. [28], an 8-core parallel computing try, Drewe et al. [6] investigated the relationship between proxi-
setup was used to simulate a left ventricle (LV) and systemic mal neck and iliac bifurcation angle in idealized geometries with
circulation model and the authors reported each simulation took specific bifurcation angles to elucidated their role in rupture risk
approximately 168 h, or roughly a week, to complete. Voβ et al. and the need of treatment. Stergiou et al. [35] used FSI to exam-
[29] used a 4-core setup to model cranial aneurysm rupture, re- ine how hematocrit affects AAA progression, finding it doesn’t af-
porting 30 h of runtime per simulation and a study by Votta et al. fect von mises stress (VMS) magnitude but does affect wall shear
[21] leveraged a 12-core setup to model cardiac valve movement, stress (WSS). Using patient specific, fully-coupled ALE FSI models
reporting 6 h run time per simulation. Simulation runtime is with geometries and boundary conditions derived from 4D MRI
variable, but in all the cases referenced the time required to com- images, Campobasso et al. [36] studied the effect of aortic stiffness
plete runs limited the number of scenarios reasonably modeled. and hypertension on peak wall stress in ascending thoracic aortic
These constraints are being mitigated in two ways. First, there is aneurysms (TAA). The authors found both hypertension and wall
a substantial body of work being completed to create more effi- stiffness increase peak wall stress, explaining clinical observations
cient computational methods that require less resources and have of increased rupture risk in certain patient populations. Ong et al.
higher accuracy. Second, the available computational resources [37] used Two-Way FSI to examine intra-luminal thrombus (ILT)
continue to increase. Researchers leverage small clusters in the lab formation in TAA, finding that high vessel curvature can cause ILT
and others utilize larger clusters that they have available on their because high shear stress activates the blood recirculating in the
campus. Cloud computing is also becoming more common because aneurysm sac, causing thrombus formation. These studies show the
it allows researchers to access to the cutting-edge computational utility of FSI in studying the effect of biomechanical markers on AA
resources that can be scaled without needing to purchase new progression and rupture risk.
hardware each year. Several reviewed studies focused on the development of non-
invasive diagnostic techniques. Polanczyk et al. [38] and Wang
7.1. Clinical relevance and applications in cardiovascular medicine et al. [39] used FSI and patient specific imaging data to reconstruct
wall strain and determine wall biomechanical properties such as
7.1.1. Vascular aneurysms WSS. Vessel strain cannot be measured directly through imaging
Many of the early FSI studies in cardiovascular engineering but is important to rupture risk. Sazonov et al. [30] published
and medicine focused on flow and structural stress in specific a study that demonstrates a non-invasive method of determining
pathologies, like vascular aneurysms. FSI is well suited for inves- the location and severity of AA by analyzing the pulse waveforms.
tigating both the structural and fluid dynamics because the struc- Similarly, Khamdaeng et al. [40] looked at how the propagation
tural and fluid domains are highly dependent on each other. A of the pressure pulse wave through the circulatory system is af-
method to predict rupture risk in patient specific anatomies and fected by an aneurysm in the aortic arch. Nikolov et al. [41], Yeh
flow conditions will allow surgeons to make informed decisions et al. [42] and Vinoth et al. [43] also published papers develop-
about treatment strategy, urgency, and timing. The rupture of aor- ing AA models. These studies collectively work toward creating a
tic aneurysms (AA) is a global leading cause of death, with more non-invasive diagnostic method for predicting aneurysm progres-
than 1,0 0 0,0 0 0 deaths per year [30]. Cerebral aneurysm are expe- sion, rupture risk, and other complications such as ILT.
rienced by 3–6% of people and have a mortality rate of 75% [4]. AA model development and validation have been studied by
It should be noted that FSI models that highlight the de- several groups. Mesri et al. [44] used a patient specific AAA ge-
formation and strain of biological tissues, like those addressing ometry in a Two-Way ALE FSI model to determine the effects of
aneurysm, typically rely on complex constitutive material mod- wall thickness and material model isotropy on the results, finding
els that are difficult to develop and validate. A few representa- models with isotropic material properties and uniform wall thick-
tive studies that highlight these types of models are published by ness underestimated wall shear stress compared to anisotropic
Bianchi et al. [31], Gilmanov et al. [32], and Balzani et al. [33]. Bi- variable wall models. The study demonstrates the importance of
ological tissue are inherently anisotropic and nonlinear, and their proper model development to mimic true physiology. Lin et al.
material properties are dependent on the patient, location/type [45] examined the results of CFD, FEA, and FSI studies on AA and
of vasculature and disease state. For this reason, material mod- again showed how model development greatly affects results. CFD
els need to be developed and adapted to the specific problem alone was unable to resolve vortices and the validity of FEA alone
being assessed and this can be challenging. Once developed and was dependent on knowing the pressure profile in aneurysm sac.
validated, material models that capture deformation characteris- Mendez et al. [1] also examined the effect of modeling technique
tics and interaction of their underlying components have higher for ascending thoracic aorta aneurysms, studying a system with
accuracy and provide a better representation of the physiological both a normal tricuspid aortic valve (AV) and a pathologic bicus-
response. The cited papers focus on development and validation pid AV. Their results showed the CFD and FEA have validity com-
of several vascular constitutive models, including the Saint-Venant pared to Two-Way FSI for certain parameters. These studies all
and May–Newman and Yin models [32] and a model based on his- demonstrated the importance of modeling methodology and mate-
tology and the cells and proteins underlying vascular material re- rial model and showed FSI should be the baseline when studying
sponses [31]. AA but that CFD and FEA alone can be useful in limited situations.
These should be investigated for each model prior to their inde-
7.1.2. Abdominal aneurysms pendent application.
FSI is widely used to study the structural and fluid dynamics of
abdominal aortic aneurysms (AAA). Maximum transverse diameter 7.1.3. Cerebral and other aneurysms
of AAA (DAAA) is the metric commonly used to determine surgical The other commonly studied aneurysm type occurs in the cere-
intervention, but the metric has proved inadequate. Canchi et al. bral arteries. Jahed et al. [4] published a paper studying pres-
[34] used two patient specific AAA geometries to study effects of sure and WSS in the Circle of Willis, an integral blood path for
6 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

the brain, using two patient specific boundary conditions and ge- Two-Way ALE FSI to study the effects of angle between mitral
ometries from CT images, one geometry from a 10 year old with and aortic orifices on wall shear stress, pressure and flow in the
a 29 × 30 mm aneurysm in the Basilar artery, and one geometry LV. Bahraseman et al. [57] used Two-Way ALE FSI to create a
from a 70 year old with a 7 × 6.9 mm aneurysm in the left mid- model of LV stroke work, validated against literature, that can
dle cerebral artery. The authors found WSS, tension, and rupture predict stroke work at different heart rates to determine the
risk is greatest in the aneurysm neck and not fundus in their mod- effects of exercise on the heart. Deng et al. [58–60] published a
els and concluded the parameters are highly dependent on patient series of studies focused on modeling the effects of hypertrophic
specific geometry. Antonov et al. [46] used One-Way FSI to study cardiomyopathy (HCM). The first paper establishes a model that
the relationship between location, severity and rupture risk in ge- investigates systolic anterior motion (SAM) in patients with HCM
ometries from sixty-nine middle cerebral aneurysm patients, de- and then subsequent papers investigated specific elements of
termining geometric parameters that reduce rupture risk. The re- SAM in HCM for patients receiving a septal myectomy surgery.
sults of Valencia et al. [47] support the use of One-Way FSI in Govindarajan et al. [61] and Khalafvand et al. [62] also created
cerebral aneurysms. The group also studied the effects of hyper- LV FSI models with native valves. FSI is well suited to model LV
tension and pressure gradient on cerebral aneurysms, finding hy- dynamics because the significant interplay between the fluid and
pertension increases WSS and displacement but does not affect structural domains is fundamental to LV function.
overall hemodynamics [48]. Hajirayat et al. [49] used FSI to study There were also studies focused on modeling of the aortic valve,
the effect of cerebral aneurysm neck shape on rupture risk and considering the effects on the downstream large arteries. Tango
hemodynamics, finding a circular neck case resulted in a greater et al. [63] published a paper focused on aortic valve dynamics, first
WSS than the elliptical neck and had a 40.8% increase in the like- tailoring the FSI model to agree with experimental particle image
lihood of rupture. These studies collectively can be used to char- velocimetry (PIV) measurements and then extending the model to
acterize cerebral aneurysms and make informed clinical decisions. overcome the shortcomings of the experimental setup. Gilmanov
Shamloo et al. [50] published a study modeling treatment strate- et al. [64] worked on an aortic valve model, with a focus on im-
gies for Circle of Willis aneurysms, reviewed in the surgical mod- proving the material model used for the valve and Hasan et al.
eling subsection. In addition to the work published on aortic and [65] successfully created a patient specific IB based FSI model of
cerebral aneurysms, Nikolov et al. [51] published a study modeling aortic root and ascending aorta to be used to plan valve repair
an aneurysm in the carotid artery and Sharzehee et al. [52] pub- or replacement surgeries. Fedele et al. [66] also created a model
lished a study investigating the effects of tortuosity in carotid of the aorta that included the valve. Cao et al. [67] published
and aortic aneurysms. These studies demonstrate how FSI can be a study investigating the contribution of coronary blood flow to
used to study complex patient specific flows to inform clinical de- aortic valve dynamics, finding coronary flow decreased vortex de-
cisions and further our understanding of the dynamics of these velopment and altered WSS distributions. Similarly, Mohammadi
anatomies. et al. [68–70] published three papers that study the contribution
In a different way of leveraging FSI, Helthuis et al. [53] stud- of coronary arteries on the aortic valve, aorta and aortic root.
ied the efficacy of partial occlusion for treating cerebral aneurysm. Sigüenza et al. [71] used FSI as part of a larger study that inves-
Two patients were imaged with MRI before and after bypass tigated the effect of pulsatility on aortic valve flow using PIV. Stu-
surgery, one with proximal partial occlusion and one with distal pak et al. [72] demonstrated the importance of selecting the proper
partial occlusion. FSI was performed on pre- and post-operative ge- turbulence model when completing FSI on aortic valve flow, com-
ometries to determine the hemodynamics and wall structural dy- paring results from simulations that did not leverage a turbulence
namics. Results were validated against angiograms and CT images model with common turbulence models, including k-epsilon and
and it was found proximal occlusion showed decrease in WSS and several k-omega based models. Turbulence models are often re-
velocity in whole aneurysm, whereas distal occlusion only saw re- quired to numerically solve the Navier Stokes equations and simu-
ductions in the aneurysm dome. lations that do not use a turbulence model are more robust but
require significantly more computational resources. The authors
8. Cardiac valve dynamics found the k-omega based shear stress transport model produced
results most similar to the turbulence model free simulations.
FSI is inherently suited for the modeling of the dynamics of Several reviewed studies investigated the causes of mitral valve
native and prosthetic heart valves since movement is dependent prolapse and regurgitation. Toma et al. [73,74] published two pa-
on the pressures and flows in the chambers of the heart and pers focused on the chordae structure of the mitral valve, first
the hemodynamics are in turn dependent on valve state. Fluid– demonstrating the importance of including the chordae in the
structure interaction, particularly fully coupled Two-Way FSI, is model, and then investigated 51 possible chordae rupture points
used to model movement of the heart valves, blood flow, and how to determine the effect of chordae diameter and strain on regur-
they affect each other in a transient system. Two-Way FSI has be- gitant orifice area. Khodaei et al. [75] used a Two-Way ALE FSI
come the modeling standard for the study of cardiac valve dynam- model to compare stresses on mitral valve leaflet and chordae in
ics. The IB method is best suited because the thin valve leaflets healthy and prolapse patients, finding an increase in stress on the
undergo significant deformation within the fluid domain and the interior and posterior leaflets in patients with prolapse, and an in-
remeshing required to implement a mesh conforming method is crease in chordae tension in these regions when chordae rupture
computationally intensive and onerous [15,16]. occurs elsewhere. Caballero et al. [76] developed 8 FSI models, one
The dynamics of the left side of the heart, specifically the healthy and seven rupture models, to study the effects of chordae
LV and the mitral and aortic valves, were modeled using FSI rupture on mitral valve regurgitation, finding the basal/strut chor-
in several reviewed papers. Gao et al. [54] developed a LV IB dae bore most of the load. Chordae rupture reduced basal/strut
based Two-Way FSI model that models a complete cardiac cycle tension but increased the load elsewhere in the leaflet. Feng et al.
and includes valve features, LV contraction, non-linear structural [77] created an IB FSI model of mitral valve chordae rupture, com-
mechanics of valves and the interaction of the fluid, ventricular paring three methods of incorporation into the structural model,
wall and valve. Similarly, Mao et al. [55] created a fully-coupled finding their “pseudo-fiber” method yields the greatest accuracy at
FSI model of the LV, including both the mitral and aortic valves, the least computational cost. Hassani et al. [78] also studied mitral
over a full cardiac cycle. In both of these studies, the authors valve dynamics, focusing on poor leaflet coaptation and regurgi-
validated their model against in vivo data. Arefin et al. [56] used tation. The valve and underlying structures present a complicated
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 7

system but a validated valve pathology model will help clinicians in the bulk flow than hinge or leakage flow due to higher veloc-
make informed treatment decisions. ities, motivating the second paper, in which the authors studied
FSI is also being used to study the effects and dynamics of con- bulk flow platelet activation during systole for bioprosthetic and
genital bicuspid aortic valves (BAV), the most common congenital mechanical heart valves, finding early in systole the bioprosthetic
valve defect. In BAV, two of the three leaflets of the aortic valve valve have slightly higher platelet activation but at the end of sys-
are fused, resulting in a bicuspid aortic valve. The classification of tole the mechanical valve has much higher platelet activation in
BAVs are based on their cusp fusion pattern, with type-0 indicat- all tested activation models. These results motivate the continued
ing leaflets of equal size and type-1 indicating the leaflets are of development of bioprosthetic valves that have dynamics more like
unequal size. Cao et al. [25,79] published two studies in this area, native valves. Abu Bakar et al. [91] used a PIV validated FSI model
first determining the WSS profile and leaflet deformation of nor- to compare velocity and vorticity in a bi-leaflet heart valve. There
mal tricuspid aortic valves (TAV), type-0 BAV and type-1 BAVs and is a significant need for artificial heart valves, and these studies
then investigating aortic dilation in TAVs and type-1 BAVs, find- show how FSI can be leveraged to improve patient outcomes.
ing type-1 BAVs result in abnormal aortic hemodynamics in ar- Transcatheter aortic valve replacement (TAVR) is commonly
eas with high dilation risk and indicating WSS metrics should be used clinically to replace pathologic aortic valves using a catheter
considered when predicting effects of valve anatomy. Pasta et al. to implant the prosthetic valve. It is a popular intervention because
[80] also studied aortic dilation in BAV patients, comparing the it does not require opening the chest cavity. Many of the TAVRs
results of 21 BAV and 13 tricuspid aortic valve (TAV) geometries, currently on the market were not designed to be implanted us-
finding the structural and hemodynamics are different in the two ing a catheter, and the resultant valve crimping can affect hemo-
categories. Lavon et al. [81] studied the effect of nonfused cusp dynamics [92] and FSI can be used to design and study new TAVR
(NFC) angle in type-1 BAV patients using a fully-coupled FSI model, designs that mitigate these effects. Ghosh et al. [92] studied a new
finding cusp angle affected mechanical behavior, eccentric jet di- TAVR design using FSI to compare their design to a surgical aor-
rection and fluid/structural stresses. Their results highlighted why tic valve replacement (SAVR), showing increased leaflet opening
early failure is observed in BAVs with small NFC angles and why a (42%) and flow (27%) compared to the SAVR design. The design also
large NFC is preferable in suture annuloplasty BAV repair surgery. had WSS and mechanical stresses similar to the SAVR. Borowski
Liu et al. [82] used FSI to study circumferential WSS in BAV and et al. [2] compared two models, one that only considered structural
TAV patients due to helical flow in the aorta. These studies rep- valve dynamics using FEA and the other that incorporated fluid
resent examples of how FSI can be used to study patient specific dynamics into an FSI model. The FEA model alone did not prop-
pathologies to determine the consequences and plan treatment. erly model valve dynamics, showing the leaflets opening at a rate
FSI has also been used to study the effects of heart valve steno- five times that of the FSI model, demonstrating the utility of FSI.
sis. Sadeghpour et al. [83] studied flow through a stenotic aortic Kandail et al. [93] used FSI to the study of ideal placement of the
valve, finding the low orifice area that results from incomplete aor- CoreValve TAVR, comparing the hemodynamics in the ascending
tic valve opening causes a fluid jet with high blood stresses. Olcay aorta and coronary arteries for annular and supra-annular place-
et al. [84] also investigated the jet leaving the LV in patients with ments, finding the supra-annular placement was superior at peak
aortic valve stenosis using FSI with Lagrangian Coherent Structures. systole because it produced a centered jet that impinged further
Finally, in a slightly different type of native valve study, Soifer et al. downstream than the annular position. In a similar study, Vahid-
[85] used FSI to study pathologic venous valves, focusing on the khan et al. [94] investigated flow stasis, which can lead to leaflet
effects of valve stiffening and failure on downstream vessels. Their thrombosis, with intra-annular and supra-annular TAVR position-
results can be used to understand valve failure and retrograde flow ing, finding the intra-annular position limits valve opening and can
and to design prosthetic valves. Valve pathologies and stenosis can cause areas of stagnation that can lead to thrombosis. TAVR is a
lead to negative clinical outcomes and FSI is a great tool to nonin- significant improvement, and FSI is being leveraged to develop and
vasively study the consequences. improve the devices and surgical techniques.
Nestola et al. [95] used FSI to compare stented and stent-
9. Prosthetic valve dynamics less aortic valves, investigating aortic root displacement and in-
ternal stress for three patient specific geometries. Each modeled
Prosthetic heart valve dynamics are also well suited for study as a healthy case, stented valve, and stentless valve, and finding
with FSI due to the interplay between structure and fluid. Valves the stentless configurations produced results most similar to the
constructed from a fiber reinforced polymer matrix are the most healthy case. Wang et al. [16] studied the effect of suture den-
common prosthetic valve type. Avanzini [86] used FSI to study sity on valve performance using an IB based FSI study and an FEA
the role of fiber orientation and fiber stiffness in the opening of study. The FSI study outperformed the FEA study and showed su-
prosthetic trileaflet heart valves, finding both parameters affect the ture density had a significant impact on principle stress and shear
transvalvular pressure gradient and valve opening. Circumferential stress, although they showed opposite trends.
fibers reduce the effective orifice area, as does increases in matrix Tissue engineered heart valves require additional structural
stiffness. Gharaie et al. [87] studied the non-linear deformation of support when they will be utilized on the left side of the heart
polymeric aortic valves using an ALE based Two-Way FSI model where pressure is high. Sodhani et al. [15] developed a FSI model
that was validated with in vitro benchtop testing, finding a max- to recreate in vitro testing of a scaffold reinforced tissue engi-
imum of 15% error between in silico and in vitro results. Lurahi neered heart valve. Their model produced results within 9% of
et al. [88] also validated a polymeric heart valve model, compar- the benchtop results during diastole, supporting the use of FSI
ing FSI and FEA results to benchtop testing and finding FSI pro- to develop prosthetic heart valves. Chen et al. [96] studied bio-
duced max orifice opening area results within 5% of the benchtop prosthetic aortic valve dynamics using an IB FSI model and Xu
results, whereas the FEA model deviated by 46.5%. These studies et al. [97] presented a framework for developing bioprosthetic
can be used to design and tailor reliable heart valves to patient heart valves based on a hybrid ALE-immersogeometric FSI model.
needs. Hedayat et al. [89,90] published two FSI based studies that
investigated mechanical heart valves. The first looked at the effect 10. Cardiovascular stents
of leaflet gap size on platelet activation, showing a larger gap had
higher total activation but also improved washout ability due to in- Unlike other areas of cardiovascular engineering reviewed, One-
creased flow velocity. They also observed higher platelet activation Way FSI is common when studying stents because the fluid pres-
8 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

sure does not have a significant an impact on stent deformation, to quantify the required suture strength. Tubaldi et al. [109] also
meaning the structural domain can be modeled first creating a investigated Dacron grafts, specifically examining effects of pul-
stented geometry that can then be used in CFD studies. Putra et al. satile flow and finding abnormal and unwanted radial oscillations
[98] employed One-Way FSI in a stent optimization study that used at high heart rate. Rahmani et al. [110] used FSI to investigate AA
FEA to model stent deployment and CFD to model the fluid effects repair with synthetic biomaterial grafts made of polyurethane, sil-
of the stent geometry and vessel deformation. Their optimization icone, PTFE and Dacron, finding polyurethane and silicone resulted
algorithm aimed to reduce areas of low WSS and produced an op- in greater displacements and lower velocities than PTFE or Dacron,
timized stent design with improved hemodynamics. Simão et al. demonstrating again the effect of graft material on stress profiles
[99,100] published two papers, each focused on evaluating new and hemodynamics. Georgakarakos et al. [111] used FSI to examine
stent designs using One-Way FSI, evaluating WSS and flow distur- effect of main body-to-iliac limb length ratios in aortic endografts,
bances, and showing the utility of One-Way FSI in stent design. examining pressure drop across the graft, helicity in the aorta,
The studies aimed at reducing the adverse effects on the wall that and WSS for two different surgical orientations. The results can
may lead to restenosis. Lui et al. [101] used FSI to predict WSS on a be used by surgeons to plan repair technique and the preferred
degradable stent, and the results were incorporated into a bench- limb ratio based on patient geometries. Wang et al. [112] stud-
top model that simulated degradation. ied multiple overlapping uncovered stents in complex AA with
Pulse wave propagation (PWP) is integral to circulatory system. side branch problems using a 1-way FSI model, showing stent
The large arteries of the body are elastic and aid in blood flow deployment created structural stress concentrations that increased
by responding to the cyclic pressure created by the LV and stents linearly with additional stent deployments. The study also showed
are thought to contribute to a reduction in PWP. Frecentese et al. that the pattern of overlapping stents did not influence WSS or
[102] used FSI to study PWP in stented arteries to determine if the pressure within sac, thus providing evidence to clinicians to focus
stent was expected to cause wave reflection, testing many con- less on pattern of overlap.
figurations and designs to determine if multiple small stents or Patients born with single ventricle congenital abnormalities re-
one large stent is better in areas of high stenosis, finding stent quire a series of surgeries after birth that result in the creation
geometry has a significant impact on the location of stop-bands of the total cavopulmonary connection (TCPC), where the supe-
that prevent PWP. The authors found the crosslinking of stent rior and inferior vena cavae drain directly into the pulmonary ar-
coils increases the number and width of stop-bands whereas plac- teries. The TCPC has low pulsatility and pressure that can lead
ing multiple stents with a finite distance between them results to long-term complications for these patients. Rajabzadeh-Oghaz
in fewer stop-bands. The results begin to illuminate why certain et al. [113] used FSI to develop a new approach to introduce pul-
vessels respond better to stenting than others and how to orient satility to the TCPC through a stenosed pulmonary artery, com-
stents to aid PWP. Papathanasiou et al. [103] also studied the effect paring an MRI based T-shaped graft with a computer-generated Y-
of stenting on PWP, investigating one and two stent configurations shaped graft and finding both introduced similar pulsatility, how-
and finding the single stent only induced wave reflection under ever, less energy was lost in the Y-shaped graft.
pathologic conditions. The two-stent design, however, could pro- The use of FSI to assist with surgical planning will increase in
duce wave reflection under normal physiological conditions. The frequency in the future and become the gold standard. The quality
effects of stenting on PWP should be further studied. of patient specific imaging and the availability of computing power
are both increasing. As the methods become validated surgeons
11. Surgical modeling will gain confidence in the technique and leverage the methods to
improve clinical outcomes.
FSI is being leveraged to model surgical tools and techniques
to treat cardiovascular pathologies. It is well suited because it can 12. Mechanical circulatory assist
model the complete fluid and structural physics of the complex cir-
culatory system and will yield the most clinically relevant com- FSI is being used to model the downstream and local vascular
putational results. Aspiration thrombectomy is a technique used response of mechanical circulatory assist devices. These are often
to treat cerebral artery blockages. Chitsaz et al. [104] used FSI One-Way FSI studies because the rigid impeller blades will not be
to model the performance of two novel aspiration catheters in a deformed due to fluid pressure and the additional computational
experimentally validated patient specific model. Catheter perfor- resources required to complete a Two-Way study are unnecessary.
mance was determined by evaluating the aspiration ratio of the Two-Way FSI is being leveraged in the modeling of pulsatile pneu-
clot, aspiration time, number of free fragments, WSS, and extracted matic VADs because structural deformation is the foundation the
volume of the blood. Guerciotti et al. [105] used FSI to assess and device mechanism of action. Caimi et al. [114] published an Two-
compare arterial and venous coronary artery bypass grafts, finding Way FSI study that modeled the Penn State 12cc pneumatic pedi-
venous grafts result in more disturbed flow and higher WSS at site atric pulsatile VAD, specifically trying to reproduce the interplay
of anastomosis when compared to the arterial graft. between the diaphragm, blood chamber and pneumatic actuator.
The study of aortic aneurysm repair with FSI was found in They were able to study and optimize pneumatic pulsatile VADs.
several of the reviewed studies. A paper published by van Bakel Obidowski et al. [115] used FSI to find areas of flow stagnation in
et al. [106] used eight patient specific geometries in a series of a pulsatile VAD. The authors then successfully validated the model
FSI models that investigated the effect on LV stroke work of using with PIV measurements.
a graft that is stiffer than the native aorta during an AA repair, Alizadeh et al. [116] used FSI to study the force on the aor-
finding the graft resulted in a 25% increase in LV stroke work and tic valve caused by the AVICENA VAD, which is wrapped around
an increase in LV mass. Other authors have studied the effects of the aorta, enhancing flow and pressure through counterpulsation.
more compliant graft materials. Jayendiran et al. [107,108] pub- Six sets of displacement curves and one set of boundary condi-
lished two papers that use FSI to examine AA repair techniques. In tions were applied to determine which set of operating condi-
the first paper, FSI models were used to compare four stents and tions would decrease force on the aortic valve while increasing
a graft repair, finding stent and/or graft material and geometry power generation. The initial study was expanded by Rahmani
greatly affect hemodynamics. The second paper used FSI to study et al. [117] and 12 displacements and 3 sets of boundary condi-
AA repair with a Dacron graft, finding AA repair with Dacon yields tions were applied to the model. The three sets of boundary con-
similar stress profiles to healthy aorta. Additionally, they were able ditions represented a healthy patient, a hypertensive patient and a
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 9

hypotensive patient. They again were able to ascertain which set ity. The FSI model was configured to match PIV results and then
of conditions increase power generation while minimizing aortic was expanded to investigate parameters not easily probed with PIV
valve force. Rahmani et al. [118,119] also published two additional alone. In a another study, Luc et al. [128] investigated the effect of
FSI based studies that continued development of balloon based continuous flow VADs on cerebrospinal fluid flow.
VADs that wrap the aorta. In the first paper the authors modeled
a single balloon system made of three different potential materi- 13. Atherosclerosis and stenosis
als and seven different material thicknesses. The FSI model was
extended into a second paper with a VAD design based on two 13.1. General atherosclerosis and stenosis modeling
balloons that wrap the aorta, investigating two inflation/deflation
schemes, one aimed at producing continuous flow and the other Atherosclerosis is often studied with FSI because atherosclerotic
pulsatile flow. The authors analyzed flow in the aorta and down- plaques affect hemodynamics and the local hemodynamics lead to
stream arteries. This is an example of how FSI is being leveraged in plaque initiation, growth and potentially rupture. Modeling of the
VAD development, both from a device standpoint and with respect two interconnected systems requires the use of FSI. Chhai et al.
to operating conditions. [129] used FSI to investigate factors contributing to atherosclerotic
Bakir et al. [120] examined the use of a continuous flow LVAD plaque rupture, examining at the effects of longitudinal asymme-
during heart failure in a fully coupled biventricular model that try, finding maximum cap stress, WSS and rupture risk increased
considers electrical activation, myocardial mechanics and hemody- with increasing plaque asymmetry. Tang et al. [130] investigated
namics. The model demonstrated a shape shift in the LV pressure- the effect of plaque cap inflammation on rupture using patient
volume curve after the LVAD. At high rotational speeds, aortic ejec- specific FSI models. The authors simulated plaque weakening due
tion ceased, LV volume decreased and after multiple heart cycles to inflammation, under normal conditions and with a thin cap and
the LV collapsed. The model can be used in the development of hypertension. They found cap inflammation leads to large strain
LVAD control mechanisms, specifically to prevent LV collapse. In conditions, especially when cap is thin and patient hypertensive.
another study, Bhat et al. [121] used a patient specific FSI model He et al. [5] investigated the effects of early atherosclerosis
to try to determine why the Heartmate II has a lower incidence of on neighboring vessels, finding WSS is lower and cyclic strain
stroke than the Heartware HVAD, focusing on the effect of outflow in higher in arteries near atherosclerotic vessels and showing
angle and diameter, parameters that are different in the two VADs. atherosclerosis has a negative effect on adjacent healthy vessels,
They found a 10 mm diameter outflow graft is more sensitive to contributing to disease spread and supporting the need for early
the outflow angle than a 14 mm diameter graft. These results in- diagnosis. Kafi et al. [131] published a study comparing Newtonian
form the parameters to consider when developing VAD technol- and non-Newtonian fluid formulations in atherosclerotic arteries
ogy. In a similar study, Chiu et al. [27] completed patient specific and demonstrating the importance of wall rigidity on stress pro-
FSI studies on the Heartmate II and HeartAssist 5 to determine file. Bahrami et al. [132] modeled non-Newtonian flow through the
the effects of inflow cannula angle, VAD orientation and outflow left coronary bifurcation using ALE Two-Way FSI to investigate the
graft angle and determining the optimal configuration. Liao et al. onset of atherosclerosis in an idealized geometry, considering both
[122,123] published two studies focused on elements of the contin- normal blood pressure and hypertension. Kallekar et al. [26] also
uous flow Heartware HVAD. In the first paper, the authors studied investigated the effect of wall model on FSI results of atheroscle-
the effect of speed modulation on intraventricular flow dynamics rosis, comparing Rigid, Linear Elastic, neoHookean, Mooney-Rivlin
by modeling a dilated LV supported with the Heartware HVAD at and Holzapfel material models under steady state and pulsatile
constant speed, and with sinusoidal copulsation and counterpulsa- conditions, again showing results are highly depended on wall
tion. After assessing thrombotic potential, the authors found copul- model used and demonstrating the importance of FSI model for-
sation to be a superior operation mode because it reduced resi- mulation. Nejad et al. [133] created a fully coupled ALE FSI model
dence time, and improved washout and pulsatility index at outflow of pulsatile stenotic vessel flow that considered viscoelastic and
compared to other modes. In the second paper, Liao [123] assessed elastic wall properties, atherosclerosis, and high hematocrit. Pereira
the effects of inflow canula length in the Heartware HVAD in pa- et al. [134] created an IB FSI atherosclerosis model, considering the
tients with normal apical thicknesses and apical hypertrophy, ex- particulate nature of blood and validated it against the literature.
amining pump washout, residence time, flow stagnation and pul- Gholipour et al. [135,136] developed an FSI model of stenotic
satility index. The authors found a longer insertion length relative blood flow in the coronary arteries to predict heart attack. The first
to apical thickness reduced thrombotic potential by increasing LV study considered varying levels of stenosis in a physiologically rel-
washout and decreasing residence time. Sonntag et al. [124] com- evant model that included pulsatile flow, realistic vessel material
pleted a patient specific lump parameter and FSI study to fit the properties, heart contraction and motion, and plaque properties in-
continuous flow EVAHEART 2 VAD with a double cuff tipless in- cluding the lipid core. In the second paper, the model was applied
flow canula to determine optimal length during both full and par- to the coronary artery bifurcation. These models can be used to
tial VAD support. Zhang et al. [125] investigated partial support predict heart attack given a patient specific geometry and physio-
configurations of the BJUT-II VAD, comparing intraventricular flow logical conditions. Javadzadegan et al. [137] studied the effects of
during heart failure, and with partial and full VAD support by as- degree of freedom movement, creating 25 coronary artery mod-
sessing WSS, oscillatory shear index (OSI) and residence time. The els, each with a distinct set of support points. It was found that at
results indicated partial support has the greatest benefit on intra- low levels of stenosis, greater freedom of movement greatly affects
ventricular flow. Collectively these show the utility of testing de- WSS and flow recirculation, but these effects decreased as level
sign parameters and modes of operation in a virtual setting. of stenosis increased. Meza et al. [138] created a patient specific
Luraghi et al. [126] modeled the left chamber of the Carmat fully coupled FSI model of the left anterior descending (LAD) coro-
TAH using FSI, including displacements of biological aortic and mi- nary artery with 70% stenosis that considered transient blood flow,
tral valves and displacement of pericardium. They compared the cyclic artery bending and heart muscle contraction. The authors
results of the simulation to results of a standalone CFD simula- found the plaque reduced wall tensile strain and the inclusion of
tion, finding the FSI simulation was required to properly model myocardial contraction had a significant effect on local hemody-
the system because of the valve movement. Wappenschmidt et al. namics, specifically on periodic flow velocity and locations of shear
[127] developed a hybrid rotary piston pump, blending modern stress. Pinho et al. [139] also investigated flow in the LAD using FSI,
rotary pump and older pneumatic pumps to introduce pulsatil- leveraging statistical analysis to consider the effects of artery cross
10 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

sectional area and length, angles between branches and septum, 14. Future landscape of FSI in cardiovascular applications
and artery curvature and tortuosity. Kaewbumrung et al. [140] also
published a study investigating coronary flow in stenotic vessels. The use of FSI continues to evolve and advance in computa-
Pinho et al. [141] published a study examining the effect of arterial tional methods and to investigate clinical questions and applica-
compliance on WSS in healthy and mildly stenotic carotid arteries. tions in cardiovascular disease and treatment. These studies span
the macro-domain to cellular structures – a range of fully cou-
pled electro-mechano-fluidic challenges. Further linking these ap-
13.2. Non-invasive diagnostic proaches to hemodynamic lumped parameter models of the car-
diovascular system, patient-specific anatomic geometries, vessel
Gamage et al. [24] studied PWP in a stenotic artery using fully- wall mechanics, wall shear stresses and pulsatility. Implementation
coupled Two-Way FSI model, finding PWP could be used as a diag- for pathophysiological assessment of the microcirculation, valvu-
nostic tool because stenotic vessels showed pulse wave dampening lar and vascular disorders, remodeling implications due to disease
and high vessel displacement downstream. Samaee et al. [142] also and growth, and the impact of device intervention is also realiz-
developed an experimentally validated FSI model of PWP in arter- able. In addition, image guided modeling of virtual therapies, cou-
ies to predict WSS and circumferential stress waves that can be pled to lumped parameter, electrical analog models, patient spe-
used to predict plaque growth and vessel remodeling based on cific physiology, and cardiac magnetic resonance data to establish
endothelial cell pathology. Pakraven et al. [143] created an FSI appropriate and realistic boundary conditions is in process and is
based method for determining endothelial cell (EC) morphology on the horizon. To reach this a new level of FSI fidelity, new nu-
based on geometry of coronary artery, flow, pressure, heart rate merical methods to enhance solver partitioning and model cou-
and the mechanical properties of the wall and blood. The authors pling are required. Hybrid methods that couple IB (efficient in-
found sites prone to atherosclerosis have at least one of three teraction method) lattice Boltzmann methods (complex fluid flow)
properties: (1) low time averaged WSS, (2) high angle WSS and (3) with the smoothed point interpolation method (robust structural
high longitudinal strain. This method could be used to assess dis- approach) are advancing FSI to facilitate larger structure deforma-
ease risk in patient specific models. Abdessalem et al. [144] used tions [150]. Additional advances in coupling involve parallel time-
FSI to validate a larger study aimed at determining the location of stepping schemes and algorithms to maximize processing power,
severe stenosis based on propagation constant and instantaneous rather than continuing to use sequential algorithms [151]. We are
flow measurements at two places in the arterial tree. poised to achieve a new era of FSI capabilities so that we are
Liu et al. [145] developed an FSI model to predict atheroscle- best able to inform and improve clinical decision making. This
rotic plaque progression based on WSS and von mises stress (VMS). will translate into a greater understanding of cardiovascular dis-
The authors used twelve patient specific geometries with acute ease and into improved patient outcomes, thus saving lives.
coronary artery syndrome to model WSS, VMS, and other hemo-
dynamics parameters at baseline and at a 12-month follow-up to 15. Conclusions
identify factors that led to plaque progression. Wang et al. [3] used
FSI based on intravascular ultrasound for 9 patients to assess mor- As computational resources become more readily available and
phological and biomechanical risk factors of plaque progression, advanced, and imaging modalities continue to improve their res-
determining the optimal combination of 8 parameters to predict olution, fully coupled FSI studies will continue to become more
plaque formation. common in cardiovascular engineering. Two-Way FSI studies are
more sophisticated analyses that model complex environments and
as they become more ubiquitous the quality of the models will
13.3. LDL transport that leads to atherosclerosis and stenosis continue to improve, increasing their usefulness. In addition, val-
idation of the FSI models, using both new techniques and adapting
FSI has been used to study LDL transport into the vessel wall old ones, will build confidence in the results and increase the util-
using a set of methods similar set to those used to model drug ity of the techniques.
delivery. LDL penetration is an initial step leading to plaque ini- FSI studies are of particular use in cardiovascular engineering
tiation and contributes to plaque propagation and growth. Karami because many flexible structural systems, both native and syn-
et al. [146] used FSI to model LDL transport and WSS in a ves- thetic, interact with blood. They are used in widely varied appli-
sel with 60% stenosis, considering both a constant filtration veloc- cations and provide information that could not be easily procured
ity and a variable velocity based on local pressure. They found the through other numerical techniques or experimental testing. This
filtration velocity is dependent on geometry and is not consistent review has demonstrated the use of FSI across the field of cardio-
along the wall. Roustaei et al. [147] created a multilayer FSI model vascular engineering.
of LDL penetration and accumulation in the arterial wall, in which
the wall properties are dependent on the local hemodynamics and
WSS; they found wall distention, especially in hypertension, re- Declaration of Competing Interest
duces WSS and lowers endothelial cells resistance to LDL. Thon
et al. [148] used FSI as part of larger model of early atheroscle- None.
rosis. The model included a Non-Newtonian fluid and a hyperelas-
tic solid undergoing anisotropic growth and considered the effects Funding
of pulsatility and vessel compliance. LDL transport was modeled
with a set of coupled equations where wall permeability is stress The authors wish to acknowledge the financial support as pro-
dependent. The authors found pulsatility is crucial to model valid- vided by partial support from the Philadelphia-TriUniversity Con-
ity and aorta compliance has little effect on atherosclerosis. The sortium “Dream Team” Award, for Innovation in Pediatric Research.
model was validated in a murine model. Ashraf et al. [149] used This is a consortium of Drexel University, the Children’s Hospital of
FSI to explain low plaque formation risk in the aorta due to spiral Philadelphia, and Hebrew University in Jerusalem and the U.S. De-
flow preventing LDL accumulation. These studies demonstrate how partment of Education GAANN Interdisciplinary Collaboration and
FSI is being used to understand how LDL transport contributes to Research Enterprise (iCARE) for Healthcare fellowship award (M.
plaque initiation and progression. Hirschhorn).
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 11

Ethical approval [26] Kallekar L, Viswanath C, Anand M. Effect of wall flexibility on the deformation
during flow in a stenosed coronary artery. Fluids 2017;2:16.
[27] Chiu W-C, Girdhar G, Xenos M, Alemu Y, Soares JS, Einav S, et al. Thrombore-
Not required. sistance comparison of the HeartMate II ventricular assist device with the de-
vice thrombogenicity emulation-optimized HeartAssist 5 VAD. J Biomech Eng
References 2014;136:021014. http://biomechanical.asmedigitalcollection.asme.org/article.
aspx?doi=10.1115/1.4026254.
[28] Chen WW, Gao H, Luo XY, Hill NA. Study of cardiovascular function us-
[1] Mendez V, Di Giuseppe M, Pasta S. Comparison of hemodynamic and struc-
ing a coupled left ventricle and systemic circulation model. J Biomech
tural indices of ascending thoracic aortic aneurysm as predicted by 2-way
2016;49:2445–54. doi:10.1016/j.jbiomech.2016.03.009.
FSI, CFD rigid wall simulation and patient-specific displacement-based FEA.
[29] Voß S, Glaßer S, Hoffmann T, Beuing O, Weigand S, Jachau K, et al. Flu-
Comput Biol Med 2018;100:221–9. doi:10.1016/j.compbiomed.2018.07.013.
id–Structure simulations of a ruptured intracranial aneurysm: constant versus
[2] Borowski F, Sämann M, Pfensig S, Wüstenhagen C, Ott R, Kaule S, et al. Flu-
patient-specific wall thickness. Comput Math Methods Med 2016;2016.
id–structure interaction of heart valve dynamics in comparison to finite-ele-
[30] Sazonov I, Khir AW, Hacham WS, Boileau E, Carson JM, van Loon R, et al. A
ment analysis. Curr Dir Biomed Eng 2018;4:259–62.
novel method for non-invasively detecting the severity and location of aortic
[3] Wang L, Tang D, Maehara A, Wu Z, Yang C, Muccigrosso D, et al. Fluid–struc-
aneurysms. Biomech Model Mechanobiol 2017;16:1225–42.
ture interaction models based on patient-specific IVUS at baseline and fol-
[31] Bianchi D, Monaldo E, Gizzi A, Marino M, Filippi S, Vairo G. A FSI computa-
low-up for prediction of coronary plaque progression by morphological and
tional framework for vascular physiopathology: a novel flow-tissue multiscale
biomechanical factors: a preliminary study. J Biomech 2018;68:43–50.
strategy, 47. Elsevier Ltd; 2017. p. 25–37.
[4] Jahed M, Ghalichi F, Farhoudi M. Fluid–structure interaction of patient-spe-
[32] Gilmanov A, Stolarski H, Sotiropoulos F. Flow–structure Interaction simula-
cific Circle of Willis with aneurysm: investigation of hemodynamic parame-
tions of the aortic heart valve at physiologic conditions : the role of tissue
ters. Biomed Mater Eng 2018;29:357–68.
constitutive model. 2018;140:1–9.
[5] He F, Hua L, Gao L. Computational analysis of blood flow and wall me-
[33] Balzani D, Deparis S, Fausten S, Forti D, Heinlein A, Klawonn A, et al. Nu-
chanics in a model of early atherosclerotic artery. J Mech Sci Technol
merical modeling of fluid–structure interaction in arteries with anisotropic
2017;31:1015–20.
polyconvex hyperelastic and anisotropic viscoelastic material models at finite
[6] Drewe CJ, Parker LP, Kelsey LJ, Norman PE, Powell JT, Doyle BJ. Haemodynam-
strains. 2016;1–41.
ics and stresses in abdominal aortic aneurysms: a fluid–structure interaction
[34] Canchi T, Saxena A, Ng E, Pwee EC, Narayanan S. Application of fluid–struc-
study into the effect of proximal neck and iliac bifurcation angle. J Biomech
ture interaction methods to estimate the mechanics of rupture in asian ab-
2017;60:150–6.
dominal aortic aneurysms. BioNanoScience 2018;8:1035–44.
[7] Zhang Y, Bazilevs Y, Goswami S, Bajaj CL, Hughes TJR. Patient-specific vascular
[35] Stergiou Y, Kanaris A, Mouza A, Paras S. Fluid–Structure interaction in ab-
NURBS modeling for isogeometric analysis of blood flow. In: Proceedings of
dominal aortic aneurysms: effect of Haematocrit. Fluids 2019;4:11.
the Fifteenth International Meshing Roundtable Conference, IMR 20 06; 20 06.
[36] Campobasso R, Condemi F, Viallon M, Croisille P, Campisi S, Avril S. Evalu-
p. 73–92.
ation of peak wall stress in an ascending thoracic aortic aneurysm using fsi
[8] Bazilevs Y, Calo VM, Zhang Y, Hughes TJR. Isogeometric fluid–structure in-
simulations: effects of aortic stiffness and peripheral resistance. Cardiovasc
teraction analysis with applications to arterial blood flow. Comput Mech
Eng Technol 2018;9:707–22.
2006;38:310–22.
[37] Ong CW, Yap CH, Kabinejadian F, Nguyen YN, Cui F, Chua KJ, et al. Association
[9] Torii R, Oshima M, Kobayashi T, Takagi K, Tezduyar TE. Fluid–structure inter-
of hemodynamic behavior in the thoracic aortic aneurysm to the intraluminal
action modeling of aneurysmal conditions with high and normal blood pres-
thrombus prediction: a two-way fluid structure coupling investigation. Int J
sures. Comput Mech 2006;38:482–90.
Appl Mech 2018;10:1850035.
[10] Bazilevs Y, Gohean JR, Hughes TJR, Moser RD, Zhang Y. Patient-specific iso-
[38] Polanczyk A, Podgorski M, Polanczyk M, Veshkina N, Zbicinski I, Stefanczyk L,
geometric fluid–structure interaction analysis of thoracic aortic blood flow
et al. A novel method for describing biomechanical properties of the aortic
due to implantation of the Jarvik 20 0 0 left ventricular assist device. Comput
wall based on the three-dimensional fluid–structure interaction model. Inter-
Methods Appl Mech Eng 2009;198:3534–50. doi:10.1016/j.cma.2009.04.015.
act Cardiovasc Thorac Surg 2019;28:306–15.
[11] Tezduyar TE, Sathe S, Cragin T, Nanna B, Conklin BS, Pausewang J, et al. Mod- [39] Wang Y, Joannic D, Juillion P, Monnet A, Delassus P, Lalande A, et al. Vali-
elling of fluid–structure interactions with the space-time finite elements: ar-
dation of the strain assessment of a phantom of abdominal aortic aneurysm:
terial fluid mechanics. Int J Numer Methods Fluids 2007;54:901–22. comparison of results obtained from magnetic resonance imaging and stere-
[12] Sotiropoulos F, Yang X. Immersed boundary methods for simulating fluid– ovision measurements. J Biomech Eng 2017.
structure interaction. Prog Aerosp Sci 2014;65:1–21. doi:10.1016/j.paerosci. [40] Khamdaeng T, Terdtoon P. Regional pulse wave velocity and stress in aneurys-
2013.09.003. mal arch-shaped aorta. Biomed Mater Eng 2018;29:527–49.
[13] Hou G, Wang J, Layton A. Numerical methods for fluid–structure interaction [41] Nikolov N, Tabakova S, Radev S. Gaussian model deformation of an abdomi-
– a review. Commun Comput Phys 2012;12:337–77. nal aortic aneurysm caused by the interaction between the wall elasticity and
[14] Griffith BE, Patankar NA. Immersed methods for fluid–structure interaction; the average blood pressure. Stud Comput Intell 2019:305–16. doi:10.1007/
2020. 978- 3- 319- 97277- 0_25.
[15] Sodhani D, Reese S, Aksenov A, Soğanci S, Jockenhövel S, Mela P, et al. Flu- [42] Yeh HH, Rabkin SW, Grecov D. Hemodynamic assessments of the ascending
id–structure interaction simulation of artificial textile reinforced aortic heart thoracic aortic aneurysm using fluid–structure interaction approach. Med Biol
valve: validation with an in-vitro test. J Biomech 2018;78:52–69. Eng Comput Med Biol Eng Comput 2018;56:435–51.
[16] Wang ZC, Yuan Q, Zhu HW, Shen BS, Tang D. Computational modeling for [43] Vinoth R. Evaluation of abdominal aortic aneurysm wall stress based on
fluid—structure interaction of bioprosthetic heart valve with different suture flow induced load. Int J Mech Eng Technol 2018;9:684–8. Available from:
density: comparison with dynamic structure simulation. Int J Pattern Recognit http://www.iaeme.com/IJMET/index.asp684http://www.iaeme.com/ijmet/
Artif Intell 2017;31:1757007. issues.asp?JType=IJMET&VType=9&IType=11http://www.iaeme.com/IJMET/
[17] Benra F, Dohmen HJ, Pei J, Schuster S, Wan B. A comparison of one-way and issues.asp?JType=IJMET&VType=9&IType=11.
two-way coupling methods for numerical analysis of fluid–structure interac- [44] Mesri Y, Niazmand H, Deyranlou A. Numerical study on fluid–structure in-
tions. J Appl Math 2011;2011:1–16. teraction in a patient-specific abdominal aortic aneurysm for evaluating wall
[18] Karabelas E, Gsell MAF, Augustin CM, Marx L, Neic A, Prassl AJ, et al. Towards heterogeneity and material model effects on its rupture. J Appl Fluid Mech
a computational framework for modeling the impact of aortic coarctations 2017;10:1699–709.
upon left ventricular load. Front Physiol 2018;9:1–20. [45] Lin S, Han X, Bi Y, Ju S, Gu L. Fluid–Structure interaction in abdominal aortic
[19] Ahamed M, Atique S, Munshi M, Koiranen T. A concise description of aneurysm: effect of modeling techniques. Biomed Res Int 2017;2017:1–10.
one way and two way coupling methods for fluid–structure interaction [46] Antonov A, Kono K, Greim-Kuczewski K, Hippelheuser JE, Lauric A, Malek AM.
problems american journal of engineering research (AJER). Am J Eng Res Proximal stenosis is associated with rupture status in middle cerebral artery
2017;6:86–9. aneurysms. World Neurosurg 2018;109:e835–44. doi:10.1016/j.wneu.2017.10.
[20] Pei J, Benra F-K, Dohmen H. Application of different strategies of partitioned 108.
fluid—structure interaction simulation for a single-blade pump impeller. Proc [47] Valencia A, Burdiles P, Ignat M, Mura J, Bravo E, Rivera R, et al. Fluid struc-
Inst Mech Eng Part E J Process Mech Eng 2012;226:297–308. http://journals. tural analysis of human cerebral aneurysm using their own wall mechanical
sagepub.com/. doi:10.1177/0954408911432974. properties. Comput Math Methods Med 2013;2013:1–18.
[21] Votta E, Le TB, Stevanella M, Fusini L, Caiani EG, Redaelli A, et al. Toward [48] Valencia A, Torres F. Effects of hypertension and pressure gradient in a human
patient-specific simulations of cardiac valves: state-of-the-art and future di- cerebral aneurysm using fluid structure interaction simulations. J Mech Med
rections. J Biomech 2013;46:217–28. doi:10.1016/j.jbiomech.2012.10.026. Biol 2016;17:1750018.
[22] Borazjani I. Fluid–structure interaction, immersed boundary-finite element [49] Hajirayat K, Gholampour S, Sharifi I, Bizari D. Biomechanical simulation to
method simulations of bio-prosthetic heart valves. Comput Methods Appl compare the blood hemodynamics and cerebral aneurysm rupture risk in pa-
Mech Eng 2013;257:103–16. doi:10.1016/j.cma.2013.01.010. tients with different aneurysm necks. J Appl Mech Tech Phys 2018;58:968–74.
[23] ANSYS Academic release 15.0 meshing user’s guide. [50] Shamloo A, Nejad MA, Saeedi M. Fluid–structure interaction simulation of a
[24] Gamage PPT, Khalili F, Mansy HA. Numerical modeling of pulse wave propa- cerebral aneurysm: effects of endovascular coiling treatment and aneurysm
gation in a stenosed artery using two-way coupled fluid structure interaction wall thickening. J Mech Behav Biomed Mater 2017;74:72–83. doi:10.1016/j.
(FSI). 2018. p. 1117–25. jmbbm.2017.05.020.
[25] Cao K, Atkins SK, McNally A, Liu J, Sucosky P. Simulations of morphotype- [51] Nikolov N, Tabakova S, Radev S. BLOOD flow and elasticity of carotid artery.
dependent hemodynamics in non-dilated bicuspid aortic valve aortas. J Comptes Rendus l’AcadÉmie Bulg Des Sci 2017;70:1579–84.
Biomech 2017;50:63–70. doi:10.1016/j.jbiomech.2016.11.024.
12 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13

[52] Sharzehee M, Khalafvand SS, Han HC. Fluid–structure interaction modeling of [77] Feng L, Qi N, Gao H, Sun W, Vazquez M, Griffith BE, et al. On the chor-
aneurysmal arteries under steady-state and pulsatile blood flow: a stability dae structure and dynamic behaviour of the mitral valve. IMA J Appl Math
analysis. Comput Methods Biomech Biomed Eng 2018;21:219–31. doi:10.1080/ 2018;83:1066–91.
10255842.2018.1439478. [78] Hassani K, Karimi A, Dehghani A, Tavakoli Golpaygani A, Abdi H, Espino DM.
[53] Helthuis JHG, Bhat S, Van Doormaal TPC, Kumar RK, Van Der Zwan A. Development of a fluid–structure interaction model to simulate mitral valve
Proximal and distal occlusion of complex cerebral aneurysms-implications malcoaptation. Perfus (United Kingdom). 2018;1–6.
of flow modeling by fluid–structure interaction analysis. Oper Neurosurg [79] Cao K, Sucosky P. Computational comparison of regional stress and deforma-
2018;15:217–30. tion characteristics in tricuspid and bicuspid aortic valve leaflets. Int J Numer
[54] Gao H, Feng L, Qi N, Berry C, Griffith BE, Luo X. A coupled mitral Method Biomed Eng 2017;33:1–21.
valve—left ventricle model with fluid–structure interaction. Med Eng Phys [80] Pasta S, Gentile G, Raffa GM, Bellavia D, Chiarello G, Liotta R, et al. In silico
2017;47:128–36. shear and intramural stresses are linked to aortic valve morphology in dilated
[55] Mao W, Caballero A, McKay R, Primiano C, Sun W. Fully-coupled fluid–struc- ascending aorta. Eur J Vasc Endovasc Surg 2017;54:254–63. doi:10.1016/j.ejvs.
ture interaction simulation of the aortic and mitral valves in a realistic 3D 2017.05.016.
left ventricle model. PLoS ONE 2017;12:1–21. [81] Lavon K, Halevi R, Marom G, Ben Zekry S, Hamdan A, Joachim Schäfers H,
[56] Arefin MS. An investigation on the effects of the angles between the mi- et al. Fluid–structure interaction models of bicuspid aortic valves: the effects
tral and aortic orifice during diastolic period using FSI. Pathophysiology of nonfused cusp angles. J Biomech Eng 2017;140:031010.
2017;24:133–53. doi:10.1016/j.pathophys.2017.03.002. [82] Liu J, Shar JA, Sucosky P. Wall shear stress directional abnormalities in BAV
[57] Bahraseman HG, Languri EM, Mohssenzadeh M, Amirfakhri SJ, Shojaei H, Has- aortas: toward a new hemodynamic predictor of aortopathy? Front Physiol
sani K, et al. Applied fluid–structure interaction technique to initial insight 2018;9:1–9.
into the effect of exercise on the aortic valve stroke work. Am J Biomed Sci [83] Sadeghpour F, Fatouraee N, Navidbakhsh M. Haemodynamic of blood flow
2019;6:129–38. through stenotic aortic valve. J Med Eng Technol 2017;41:108–14.
[58] Deng L, Huang X, Zuo H, Zheng Y, Yang C, Song Y, et al. Angle of attack be- [84] Olcay AB, Amindari A, Kirkkopru K, Yalcin HC. Characterization of disturbed
tween blood flow and mitral valve leaflets in hypertrophic obstructive car- hemodynamics due to stenosed aortic jets with a Lagrangian coherent struc-
diomyopathy: an in vivo multi-patient CT-based FSI study. Comput Model Eng tures technique. J Appl Fluid Mech 2018;11:375–84.
Sci 2018;116:115–25. [85] Soifer E, Weiss D, Marom G, Einav S. The effect of pathologic venous valve
[59] Deng L, Huang X, Yang C, Song Y, Tang D. Patient-specific CT-based 3D passive on neighboring valves: fluid–structure interactions modeling. Med Biol Eng
fsi model for left ventricle in hypertrophic obstructive cardiomyopathy. Com- Comput 2017;55:991–9.
put Methods Biomech Biomed Eng 2018;21:255–63. doi:10.1080/10255842. [86] Avanzini A. Influence of leaflet’s matrix stiffness and fiber orientation on
2018.1443215. the opening dynamics of a prosthetic trileaflet heart valve. J Mech Med Biol
[60] Deng L, Huang X, Yang C, Lyu B, Duan F, Tang D, et al. Numerical simulation 2017;17:1750096.
study on systolic anterior motion of the mitral valve in hypertrophic obstruc- [87] Gharaie SH, Mosadegh B, Morsi Y. In vitro validation of a numerical simu-
tive cardiomyopathy. Int J Cardiol 2018;266:167–73. doi:10.1016/j.ijcard.2018. lation of leaflet kinematics in a polymeric aortic valve under physiological
01.062. conditions. Cardiovasc Eng Technol 2018;9:42–52.
[61] Govindarajan V, Mousel J, Udaykumar HS, Vigmostad SC, McPherson DD, [88] Luraghi G, Wu W, De Gaetano F, Rodriguez Matas JF, Moggridge GD, Ser-
Kim H, et al. Synergy between diastolic mitral valve function and left ven- rani M, et al. Evaluation of an aortic valve prosthesis: fluid–structure in-
tricular flow aids in valve closure and blood transport during systole. Sci Rep teraction or structural simulation? J Biomech 2017;58:45–51. doi:10.1016/j.
2018;8:1–14 Springer US. doi:10.1038/s41598- 018- 24469- x. jbiomech.2017.04.004.
[62] Khalafvand SS, Xu F, Westenberg J, Gijsen F, Kenjeres S. Intraventricular [89] Hedayat M, Borazjani I. Comparison of platelet activation through hinge vs
blood flow with a fully dynamic mitral valve model. Comput Biol Med bulk flow in bileaflet mechanical heart valves. J Biomech 2019;83:280–90.
2019;104:197–204. doi:10.1016/j.compbiomed.2018.11.024. doi:10.1016/j.jbiomech.2018.12.003.
[63] Tango AM, Salmonsmith J, Ducci A, Burriesci G. Validation and extension of a [90] Hedayat M, Asgharzadeh H, Borazjani I. Platelet activation of mechanical
fluid–structure interaction model of the healthy aortic valve. Cardiovasc Eng versus bioprosthetic heart valves during systole. J Biomech 2017;56:111–16.
Technol 2018;9:739–51. doi:10.1016/j.jbiomech.2017.03.002.
[64] Gilmanov A, Stolarski H, Sotiropoulos F. Non-linear rotation-free shell finite- [91] Abu Bakar H, Abas A, Mokhtar NH, Razak N, Hamid MNBA. Particle image
element models for aortic heart valves. J Biomech 2017;50:56–62. doi:10. velocimetry and finite volume method study of bi-leaflet artificial heart valve.
1016/j.jbiomech.2016.11.031. J Appl Fluid Mech 2018;11:1365–75.
[65] Hasan A, Kolahdouz EM, Enquobahrie A, Caranasos TG, Vavalle JP, Griffith BE. [92] Ghosh RP, Marom G, Rotman OM, Slepian MJ, Prabhakar S, Horner M,
Image-based immersed boundary model of the aortic root. Med Eng Phys et al. Comparative fluid–structure interaction analysis of polymeric tran-
2017;47:72–84. doi:10.1016/j.medengphy.2017.05.007. scatheter and surgical aortic valves’ hemodynamics and structural mechanics.
[66] Fedele M, Faggiano E, Dedè L, Quarteroni A. A patient-specific aortic valve J Biomech Eng 2018;140:121002.
model based on moving resistive immersed implicit surfaces. Biomech Model [93] Kandail HS, Trivedi SD, Shaikh AC, Bajwa TK, O’Hair DP, Jahangir A, et al.
Mechanobiol 2017;16:1779–803. Impact of annular and supra-annular corevalve deployment locations on
[67] Cao K, Sucosky P. Aortic valve leaflet wall shear stress characterization re- aortic and coronary artery hemodynamics. J Mech Behav Biomed Mater
visited: impact of coronary flow. Comput Methods Biomech Biomed Eng 2018;86:131–42. doi:10.1016/j.jmbbm.2018.06.032.
2017;20:468–70. doi:10.1080/10255842.2016.1244266. [94] Vahidkhah K, Azadani AN. Supra-annular valve-in-valve implantation reduces
[68] Mohammadi H, Cartier R, Mongrain R. 3D physiological model of the aortic blood stasis on the transcatheter aortic valve leaflet. J Biomech 2017;58:114–
valve incorporating small coronary arteries. Int J Numer Method Biomed Eng 22. doi:10.1016/j.jbiomech.2017.04.020.
2017;33:1–13. [95] Nestola MGC, Faggiano E, Vergara C, Lancellotti RM, Ippolito S, Antona C,
[69] Mohammadi H, Cartier R, Mongrain R. Fiber-reinforced computational model et al. Computational comparison of aortic root stresses in presence of stent-
of the aortic root incorporating thoracic aorta and coronary structures. less and stented aortic valve bio-prostheses. Comput Methods Biomech
Biomech Model Mechanobiol 2017;17:1–21. Biomed Eng 2017;20:171–81.
[70] Mohammadi H, Cartier R, Mongrain R. The impact of the aortic valve im- [96] Chen Y, Luo H. A computational study of the three-dimensional fluid–
pairment on the distant coronary arteries hemodynamics: a fluid—structure structure interaction of aortic valve. J Fluids Struct 2018;80:332–49. doi:10.
interaction study. Med Biol Eng Comput 2017;55:1859–72. 1016/j.jfluidstructs.2018.04.009.
[71] Sigüenza J, Pott D, Mendez S, Sonntag SJ, Kaufmann TAS, Steinseifer U, [97] Xu F, Morganti S, Zakerzadeh R, Kamensky D, Auricchio F, Reali A, et al. A
et al. Fluid–structure interaction of a pulsatile flow with an aortic valve framework for designing patient-specific bioprosthetic heart valves using
model: a combined experimental and numerical study. Int J Numer Method immersogeometric fluid–structure interaction analysis. Int J Numer Method
Biomed Eng 2018;34:1–19. Biomed Eng 2018;34:1–25.
[72] Stupak E, Kačianauskas R, Kačeniauskas A, Starikovičius V, Maknickas A, [98] Putra NK, Palar PS, Anzai H, Shimoyama K, Ohta M. Comparative study be-
Pacevič R, et al. The geometric model-based patient-specific simulations of tween different strut’s cross section shape on minimizing low wall shear
turbulent aortic valve flows. Arch Mech 2017;69:317–45. stress along stent vicinity via surrogate-based optimization. Adv Struct Mul-
[73] Toma M, Einstein DR, Bloodworth CH, Cochran RP, Yoganathan AP, Kunzel- tidiscip Optim 2017:2097–109.
man KS. Fluid–structure interaction and structural analyses using a compre- [99] Simão M, Ferreira JM, Mora-Rodriguez J, Fragata J, Ramos HM. Behaviour
hensive mitral valve model with 3D chordal structure. Int J Numer Method of two typical stents towards a new stent evolution. Med Biol Eng Comput
Biomed Eng 2017;33:1–12. 2017;55:1019–37.
[74] Toma M, Bloodworth CH, Pierce EL, Einstein DR, Cochran RP, Yoganathan AP, [100] Simão M, Ferreira JM, Mora-Rodriguez J, Ramos HM. Structural analysis of
et al. Fluid-Structure interaction analysis of ruptured mitral chordae two different stent configurations. Comput Methods Biomech Biomed Eng
Tendineae. Ann Biomed Eng 2017;45:619–31. 2017;20:869–83. doi:10.1080/10255842.2017.1306058.
[75] Khodaei S, Fatouraee N, Nabaei M. Numerical simulation of mitral valve pro- [101] Liu D, Hu S, Yin X, Liu J, Jia Z, Li Q. Degradation mechanism of magnesium
lapse considering the effect of left ventricle. Math Biosci 2017;285:75–80. alloy stent under simulated human micro-stress environment. Mater Sci Eng
doi:10.1016/j.mbs.2016.12.008. C 2018;84:263–70. doi:10.1016/j.msec.2017.12.001.
[76] Caballero A, Mao W, McKay R, Primiano C, Hashim S, Sun W. New insights [102] Frecentese S, Argani LP, Movchan AB, Movchan NV, Carta G, Wall ML.
into mitral heart valve prolapse after chordae rupture through fluid—structure Waves and fluid–solid interaction in stented blood vessels subject areas
interaction computational modeling. Sci Rep 2018;8:1–14. 2018.
 M. Hirschhorn, V. Tchantchaleishvili and R. Stevens et al. / Medical Engineering and Physics 78 (2020) 1–13 13

[103] Papathanasiou TK, Movchan AB, Bigoni D. Wave reflection and transmis- [127] Wappenschmidt J, Sonntag SJ, Buesen M, Gross-Hardt S, Kaufmann T,
sion in multiply stented blood vessels. Proc R Soc A Math Phys Eng Sci Schmitz-Rode T, et al. Fluid dynamics in rotary piston blood pumps. Ann
2017;473:20170015. Biomed Eng 2017;45:554–66.
[104] Chitsaz A, Nejat A, Nouri R. Three-Dimensional numerical simulations of aspi- [128] Luc JG, Pierre CA, Phan K, Vahedein YS, Liberson AS, Cornwell WK, et al. Fluid
ration process: evaluation of two penumbra aspiration catheters performance. structure interaction model analysis of cerebrospinal fluid circulation in pa-
Artif Organs 2018;42:E406–19. tients with continuous-flow left ventricular assist devices. Int J Artif Organs
[105] Guerciotti B, Vergara C, Ippolito S, Quarteroni A, Antona C, Scrofani R. A com- 2018;41:1290132.
putational fluid—structure interaction analysis of coronary Y-grafts. Med Eng [129] Chhai P, Lee JH, Rhee K. Effects of longitudinal asymmetric distribution of a
Phys 2017;47:117–27. doi:10.1016/j.medengphy.2017.05.008. lipid core on plaque wall stress. J Biomech Sci Eng 2017;12:1–10.
[106] van Bakel TMJ, Arthurs CJ, Nauta FJH, Eagle KA, van Herwaarden JA, Moll FL, [130] Tang D, Yang C, Huang S, Mani V, Zheng J, Woodard PK, et al. Cap in-
et al. Cardiac remodelling following thoracic endovascular aortic repair for flammation leads to higher plaque cap strain and lower cap stress: an
descending aortic aneurysms. Eur J Cardio-Thoracic Surg 2018;0:1–10. MRI-PET/CT-based FSI modeling approach. J Biomech 2017;50:121–9.
[107] Jayendiran R, Nour B, Ruimi A. Computational fluid–structure interaction [131] Kafi O, Khatib NE, Tiago J, Sequeira A. Numerical simulations of a
analysis of blood flow on patient-specific reconstructed aortic anatomy and 3D fluid–structure interaction model for blood flow in an atheroscle-
aneurysm treatment with Dacron graft. J Fluids Struct 2018;81:693–711. rotic artery. Math Biosci Eng 2016;14:179–93. http://www.aimsciences.org/
[108] Jayendiran R, Nour B, Ruimi A. Fluid–structure interaction (FSI) analysis of journals/displayArticlesnew.jsp?paperID=13081.
stent-graft for aortic endovascular aneurysm repair (EVAR): material and [132] Bahrami S, Norouzi M. A numerical study on hemodynamics in the left coro-
structural considerations. J Mech Behav Biomed Mater 2018;87:95–110. nary bifurcation with normal and hypertension conditions. Biomech Model
[109] Tubaldi E, Païdoussis MP, Amabili M. Nonlinear dynamics of Dacron aortic Mechanobiol 2018;17:1785–96. doi:10.1007/s10237- 018- 1056- 1.
prostheses conveying pulsatile flow. J Biomech Eng 2018;140:061004. [133] Nejad AA, Talebi Z, Cheraghali D, Shahbani-Zahiri A, Norouzi M. Pulsatile flow
[110] Rahmani S, Jarrahi A, Navidbakhsh M, Alizadeh M. Investigating the perfor- of non-Newtonian blood fluid inside stenosed arteries: investigating the ef-
mance of four specific types of material grafts and their effects on hemo- fects of viscoelastic and elastic walls, arteriosclerosis, and polycythemia dis-
dynamic patterns as well as on von Mises stresses in a grafted three-layer eases. Comput Methods Program Biomed 2018;154:109–22.
aortic model using fluid–structure interaction analysis. J Med Eng Technol [134] Pereira GC. A multiscale haemorheological computer-based model of
2017;41:630–43. doi:10.1080/03091902.2017.1382590. atherosclerosis: an in-depth investigation of erythrocytes-driven flow charac-
[111] Georgakarakos E, Xenakis A, Georgiadis GS. Estimating the hemodynamic in- teristics in atheroma development. J Biomech Eng Med Dev 2016;1:119.
fluence of variable main body-to-iliac limb length ratios in aortic endografts. [135] Gholipour A, Ghayesh MH, Zander A. Nonlinear biomechanics of bifurcated
Int Angiol 2018;37:41–5. atherosclerotic coronary arteries. Int J Eng Sci 2018;133:60–83. doi:10.1016/j.
[112] Wang S, Zhang Y, Feng J, Huang Y, Tokgoz A, Sadat U, et al. Influence of ijengsci.2018.08.003.
overlapping pattern of multiple overlapping uncovered stents on the lo- [136] Gholipour A, Ghayesh MH, Zander A, Mahajan R. International journal of en-
cal mechanical environment: a patient-specific parameter study. J Biomech gineering science three-dimensional biomechanics of coronary arteries. Int J
2017;60:188–96. doi:10.1016/j.jbiomech.2017.06.048. Eng Sci 2018;130:93–114. doi:10.1016/j.ijengsci.2018.03.002.
[113] Rajabzadeh-Oghaz H, Firoozabadi B, Saidi MS, Monjezi M, Navabi Shirazi MA, [137] Javadzadegan A, Yong ASC, Chang M, Ng MKC, Behnia M, Kritharides L.
Malakan Rad E. Pulsatile blood flow in total cavopulmonary connection: a Haemodynamic assessment of human coronary arteries is affected by de-
comparison between Y-shaped and T-shaped geometry. Med Biol Eng Comput gree of freedom of artery movement. Comput Methods Biomech Biomed Eng
2017;55:213–24. 2017;20:260–72.
[114] Caimi A, Sturla F, Good B, Vidotto M, De Ponti R, Piatti F, et al. Toward the [138] Meza D, Rubenstein DA, Yin W. A fluid–structure interaction model of the left
virtual benchmarking of pneumatic ventricular assist devices: application of coronary artery. J Biomech Eng 2018;140:121006.
a novel fluid–structure interaction-based strategy to the Penn state 12 cc de- [139] Pinho N, Castro CF, António CC, Bettencourt N, Sousa LC, Pinto SIS. Correlation
vice. J Biomech Eng 2017;139:081008. between geometric parameters of the left coronary artery and hemodynamic
[115] Obidowski D, Reorowicz P, Witkowski D, Sobczak K, Jóźwik K. Methods for descriptors of atherosclerosis: FSI and statistical study. Med Biol Eng Comput
determination of stagnation in pneumatic ventricular assist devices. Int J Artif 2019;57:715–29.
Organs 2018. [140] Kaewbumrung M, Wiwatanapataphee B, Orankitjaroen S, Siriapisith T. Nu-
[116] Alizadeh M, Rahmani S, Tehrani P. Calculating the aortic valve force and merical simulation of turbulent blood flow in the system of coronary arteries
generated power by a specific cardiac assist device (AVICENA) in different with stenosis. J Biom Biostat 2017;08.
counterpulsation. J Brazilian Soc Mech Sci Eng 2018;40:1–9. doi:10.1007/ [141] Pinho N, Bento M, Sousa LC, Pinto S, Castro CF, António CC, et al. Patient-
s40430- 018- 1208- 3. specific study of a stenosed carotid artery bifurcation using fluid–structure
[117] Rahmani S, Alizadeh M, Tehrani P, Navidbakhsh M. Performance and biome- interactive simulation. Lect Notes Comput Vis Biomech 2018;27:495–503.
chanical analysis of an intra-aortic cardiac assist device in different boundary doi:10.1007/978- 3- 319- 68195- 5.
conditions. J Mech Sci Technol 2018;32:3995–4002. [142] Samaee M, Tafazzoli-Shadpour M, Alavi H. Coupling of shear–circumfer-
[118] Rahmani S, Navidbakhsh M, Alizadeh M. Investigation of a new prototype ential stress pulses investigation through stress phase angle in FSI mod-
of multi-balloons LVAD using FSI. J Brazil Soc Mech Sci Eng 2018;40:1–15. els of stenotic artery using experimental data. Med Biol Eng Comput
doi:10.1007/s40430- 017- 0923- 5. 2017;55:1147–62.
[119] Rahmani S, Oveysi M, Heidari A, Navidbakhsh M, Alizadeh M. Numeri- [143] Pakravan HA, Saidi MS, Firoozabadi B. A multiscale approach for determining
cal modeling of a prototype cardiac assist device by implementing fluid– the morphology of endothelial cells at a coronary artery. Int J Numer Method
structure interaction. Artery Res 2018;22:24–35. doi:10.1016/j.artres.2018.01. Biomed Eng 2017;33:1–22.
002. [144] Ben AK, Ben SR. A new formula for predicting the position of severe arterial
[120] Bakir AA, Al Abed A, Stevens MC, Lovell NH, Dokos S. A multiphysics biven- stenosis. Comput Methods Biomech Biomed Eng 2017;20:1096–103. doi:10.
tricular cardiac model: simulations with a left-ventricular assist device. Front 1080/10255842.2017.1334769.
Physiol 2018;9:1–25. [145] Liu X, Wu G, Xu C, He Y, Shu L, Liu Y, et al. Prediction of coronary plaque
[121] Bhat S, Mathew J, Balakrishnan KR, Krishna Kumar R. Effect of outflow graft progression using biomechanical factors and vascular characteristics based on
size on flow in the aortic arch and cerebral blood flow in continuous flow computed tomography angiography. Comput Assist Surg 2017;22:286–94.
pumps. ASAIO J 2016;63:144–9. [146] Karami F, Hossainpour S, Ghalichi F. Numerical simulation of low-density
[122] Liao S, Wu EL, Neidlin M, Li Z, Simpson B, Gregory SD. The influence of rotary lipoprotein mass transport in human arterial stenosis – calculation of the fil-
blood pump speed modulation on the risk of intraventricular thrombosis. Ar- tration velocity. Biomed Mater Eng 2017;29:95–108.
tif Organs 2018;42:943–53. [147] Roustaei M, Nikmaneshi MR, Firoozabadi B. Simulation of low density
[123] Liao S, Neidlin M, Li Z, Simpson B, Gregory SD. Ventricular flow dynamics lipoprotein (LDL) permeation into multilayer coronary arterial wall: interac-
with varying LVAD inflow cannula lengths: in-silico evaluation in a multiscale tive effects of wall shear stress and fluid–structure interaction in hyperten-
model. J Biomech 2018;72:106–15. doi:10.1016/j.jbiomech.2018.02.038. sion. J Biomech 2017;67:114–22. doi:10.1016/j.jbiomech.2017.11.029.
[124] Sonntag SJ, Zebrowski E, Neidlin M, Hugenroth K, Benkowski R, Motomura T, [148] Thon MP, Hemmler A, Glinzer A, Mayr M, Wildgruber M, Zernecke-Madsen A,
et al. Virtual fitting and hemodynamic simulation of the EVAHEART 2 left et al. A multiphysics approach for modeling early atherosclerosis. Biomech
ventricular assist device and double-cuff tipless inflow cannula. ASAIO J Model Mechanobiol 2018;17:617–44.
2018;1. [149] Ashraf F, Cheema TA, Park CW. The impact of pulsatile spiral flow on the wall
[125] Zhang Q, Gao B, Chang Y. Computational analysis of intra-ventricular flow deformation characteristics and low-density lipoproteins accumulation in the
pattern under partial and full support of BJUT-II VAD. Med Sci Monit aorta. Appl Rheol 2018;28.
2017;23:1043–54. [150] Gilmanov A, Barker A, Stolarski H, Sotiropoulos F. Image-guided fluid–struc-
[126] Luraghi G, Wu W, De Castilla H, Rodriguez Matas JF, Dubini G, Dubuis P, ture interaction simulation of transvalvular hemodynamics: quantifying the
et al. Numerical approach to study the behavior of an artificial ventricle: flu- effects of varying aortic valve leaflet thickness. Fluids 2019;4.
id–structure interaction followed by fluid dynamics with moving boundaries. [151] Margenberg N, Richter T. Parallel time-stepping for fluid–structure interac-
Artif Organs 2018;42:E315–24. tions. 2019;1–19. arXiv:1907.01252.