Contents lists available at ScienceDirect

Computers in Biology and Medicine

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

Computational simulation of TEVAR in the ascending aorta for optimalendograft selection: A patient-specific case studyR.M. Romarowskia,∗, M. Contib, S. Morgantic, V. Grassid, M.M. Marrocco-Trischittad,S. Trimarchid,e, F. Auricchiob

a 3D and Computer Simulation Laboratory, IRCCS Policlinico San Donato, San Donato Milanese, ItalybDepartment of Civil Engineering and Architecture, University of Pavia, Pavia, ItalycDepartment of Electrical, Computer, and Biomedical Engineering, University of Pavia, Pavia, ItalydDivision of Vascular Surgery II, IRCCS Policlinico San Donato, San Donato Milanese, Italye Department of Scienze Biomediche per la Salute, University of Milan, Milan, Italy

A R T I C L E I N F O

Keywords:TEVARAscending aortaComputational simulationsFinite element analysis

A B S T R A C T

Thoracic endovascular aortic repair of the ascending aorta is becoming an option for patients considered unfit foropen surgery. Such an endovascular procedure requires careful pre-operative planning and the customization ofprosthesis design. The patient-specific tailoring of the procedure may call for dedicated tools to investigatevirtual treatment scenarios. Given such considerations, the present study shows a computational framework forchoosing and deploying stent-grafts via Finite Element Analysis, by supporting the device sizing and selection ina real case dealing with the endovascular treatment of a pseudoaneurysm. In particular, three devices withvarious lengths and materials were examined. Two off-the-shelf devices were computationally tested: onecomposed of Stainless Steel rings with a nominal length of 60 mm and another one with Nitinol rings and a distalfree flow extension, with a nominal length of 70mm. In third place, a custom-made stent-graft, also with Nitinolrings and containing both proximal and distal bare extensions with a nominal length of 75 mm, was deployed.The latter solution based on patient morphology and virtually benchmarked in this simulation framework, en-hanced the apposition to the wall by reducing the distance between the skirt and the vessel from more than 6mmto less than 2mm in the distal sealing zone. Our experience shows that in-silico simulations can help choosing theright endograft for the ascending aorta as well as the right deployment sequence. This process may also en-courage vendors to develop new devices for cases where open repair is unfeasible.

1. Introduction

Open surgical repair has been lately replaced by endovascularprocedures in order to treat thoracic aortic aneurysms, mostly in olderpatients with suitable anatomic features [1,2]. In contrast with theabdominal aorta, the peculiar biomechanical environment of the arch[3] characterized by the presence of the supra-aortic branches, makesendovascular planning still complex and challenging. The so-calledZone 0 (as defined in the aortic arch map by Ishimaru [4]), representsthe portion of the aorta ranging from the annulus to the ostium of thebrachiocephalic trunk. Recent studies [5] have shown that this area issubject to high hemodynamic displacement forces with pulsatile nature,which jeopardize terminal fixation stability leading to endoleak and/ormigration in a long term basis. However, the higher surface areacombined with a moderate angulation and negligible tortuosity of Zone

0 [6], protect the ascending aorta from the aforementioned adverseevents [7] despite of the hostile hemodynamics.

From a surgical point of view, endovascular repair of the ascendingaorta, when involves the aortic arch, requires a prophylactic reroutingof the supra-aortic branches by either surgical [7] or endovasculartechniques [6,8]. A similar concern can also be present for treatingisolated ascending lesions, in particular when the length of the as-cending aorta is smaller than 10 cm. Since such measurement re-presents the shortest thoracic standard endograft actually available inthe market, a longer device could determine neurological ischemia dueto the coverage of the supra-aortic branches. In these patients, themagnitude and the complexity of the debranching procedures representa relevant challenge associated with significant morbidity. As a con-sequence, the development and validation of standalone endovascularprocedures for the treatment of aortic disease specifically involving

https://doi.org/10.1016/j.compbiomed.2018.10.014Received 29 June 2018; Received in revised form 14 October 2018; Accepted 14 October 2018

∗ Corresponding author. via Morandi 30, San Donato Milanese, 20097, Italy.E-mail address: rodrigo.romarowski@grupposandonato.it (R.M. Romarowski).

Computers in Biology and Medicine 103 (2018) 140–147

0010-4825/ © 2018 Elsevier Ltd. All rights reserved.

T

Zone 0 represent an extremely relevant clinical issue [9,10].In the last decade several cases have reported successful treatments

of ascending aortic diseases by various stent-grafts and vascular ap-proaches [11], but no consensus exists regarding a standardized tech-nique. Most of the adopted solutions consist of stents, cuffs or stent-grafts as well as modified off-the-shelf devices to make them suitable fora correct deployment in the ascending aorta. On the one hand, suchmodifications reflect the current paucity of off-the-shelf devices suitablefor deployment in the ascending aorta; on the other hand, the attemptto tackle a number of heterogeneous anatomic characteristics, chal-lenges the effectiveness of endovascular treatment. Complications canoccur in the cardiac side, such as aortic insufficiency and myocardialinfarction [12,13] and they should be taken into consideration whenchoosing the proximal landing zone. Conversely, in the distal end stent-graft malapposition can induce occlusion of the brachiocephalic arteryand also stroke associated to malperfusion or embolization [14,15].These constraints call for a device with short length and wide diameter,features that are not met by stent-grafts designed for the descendingthoracic or abdominal aorta.

Given such considerations, successful endovascular treatment ofascending aortic diseases requires: 1) accurate patient selection, mostlywhen open surgery is unattainable; 2) dedicated device sizing sup-ported by design customization when necessary; and 3) a precise intra-operative stent-graft deployment based on a careful pre-operativeplanning.

To reach these aims, patient-specific simulation of stent-graft de-ployment has already proved to be a valuable tool in both abdominal[16–21] and descending thoracic aorta [22], but still needs to be cor-roborated for the ascending aorta. The seminal work by De Bock et al.[16] used the Finite Element method to deploy a bifurcated stent-graftin a 3D model of an abdominal aorta, showing excellent agreement withan in-vitro phantom. Further articles by Perrin and colleagues [17–19]used a morphing algorithm within a Finite Element Software in order topredict the final positioning of aorto-iliac endografts within the ab-dominal aorta, ranging from simple phantoms to more complicatedgeometries. In their studies, devices from different vendors were used,all showing a good agreement with in-vivo and in-vitro deployments.Recent work by Sanford et al. [21] demonstrated the ability of com-putational simulations to predict rotation in fenestrated EVAR(FEVAR). Our group has focused in merging virtual stent-graft de-ployment with CFD, also using FEA as a simulation tool [23]. All thesestudies used Finite Element Analysis with successful results, con-solidating the technique as the method of choice. To the best of ourknowledge, literature studies involving the ascending aorta are limitedto the work by Auricchio et al. [24], who compared post-operativeimages of endovascular repair with the corresponding numerical si-mulations. In this case virtual deployment was based on pre-operativeimages of an ascending aorta pseudoaneurysm later treated by acustom-made device, showing a strong quantitative agreement betweenthe real and virtual endografting configuration.

Other ”non-computational” tools have also been described in recentliterature in order to refine TEVAR planning, both to improve operatorskills and to choose the best device. As an example, Kendrick and col-leagues [25] showed that training in Angio Mentor Dual Slim en-dovascular simulator (Simbionix, Cleveland, OH, USA) decreased bothintervention time as well as radiation exposure. More recently, Bere-zowski et al. [26] deployed different stent-grafts into 3D-printed thor-acic aortic replicas and showed that a distal-to-proximal deployment (incontrast with the mostly used proximal-to-distal) improved the fixationin the distal landing zone. Malapposition in the distal landing zone hasmostly been underestimated during the deployment process but isknown to be a paramount factor in anterograde migration and type Ibendoleak [27].

Driven by a real clinical case, the present study proposes the patient-specific simulation of endovascular treatment of a pseudoaneurysm inthe ascending aorta. In particular, three different devices as well asvarious deployment sequences are tested to find the most suitable stent-graft configuration for the given vascular anatomy. Particular attentionis payed to the specific deployment procedure, which has a clear impacton the device apposition effectiveness.

2. Materials and methods

2.1. Patient characteristics

A 72-year-old male patient with a previous graft replacement of theascending aorta arrived at IRCCS Policlinico San Donato (San DonatoMilanese, Milan, Italy) for a routine follow-up. Contrast enhancedComputer Tomography (CT) was performed with a Siemens SOMATOMDefinition AS scanner (Siemens Medical Solutions, Erlangen, Germany).The scan covering the thorax had the following parameters: slicethickness = mm1 , pixel size = mm0.8 x mm0.8 , 160–650 mA, 120 kV.Images confirmed a mm47 dilation of the ascending aorta, distal to theend of the previous surgical graft. The presence of the pseudoaneurysmrequired treatment to exclude blood from filling the sac. However, dueto the overall condition of the patient, surgeons rejected the possibilityof an open-chest repair and endovascular treatment was the only pos-sible choice. Due to the retrospective nature of the study and the use ofanonymized data, the local ethics committee waived the need of pa-tient's consent for using imaging data in our analyses.

To retrieve the 3D surface required for the computational analysis, areconstruction of the aorta from the annulus to the isthmus includingthe root of the three supra-aortic vessels (brachiocephalic trunk, leftcommon carotid artery, and left subclavian artery) was performed withVMTK software suite [28]. An initial levelset algorithm was carried outvia colliding fronts and the zero-level domain was then transformedinto a 3D surface. Fig. 1 shows the reconstruction of the patient's aorta.

Fig. 1. Aorta showing a pseudoaneurysm in the ascending section. Left: sagittal CT slice, right: 3D reconstruction.

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

141

2.2. Device selection

To evaluate the viability of an endovascular procedure and find theoptimal stent-graft for this patient, three different devices were com-pared in this study. Both off-the-shelf and custom-made options werecomputationally tested.

Given the extent of the dilation and the required healthy neck, twooff-the-shelf devices were proposed by Cook Medical (Bloomington, IN,USA) with a sufficient length to divert blood flow from the lesion. Thefirst stent-graft was a Cook CMD-TBE-40-70 (device A herein) com-posed by three Stainless Steel rings sewed to a polyester (PET) skirt. Thenominal length of artery covered by the device was mm60 . This optionwas the result of a discussion between vascular surgeons and thecompany. The second stent-graft was a Cook CMD-ZTLP-A-40-65-1FF(device B herein) with three Nitinol rings and the skirt made of PET. Inthis case, a proximal bare ring was present to improve fixation in thecardiac end. The nominal length of artery covered by the device was

mm70 .Based on measurements from the 3D reconstruction and particularly

considering the severe angulation, a third device was proposed byadding a further bare ring for distal fixation and extending the coveredarea to mm75 (device C herein). This stent-graft was not in any vendorscatalogue and was designed as a virtual option to be tested with si-mulations according to the indications of vascular surgeons. Being amodification of design B, model C holds the same material properties.Fig. 2 shows the computational models of the three devices togetherwith details of their measurements. In all three cases, stent-graft dia-meter was mm40 and strut wire diameter was mm0.5 [29].

Due to their paramount importance in the mechanical behaviour,material properties of the devices were carefully reproduced from lit-erature, analysing similar endovascular procedures. Stainless Steel wascharacterized as an elastoplastic model with isotropic hardening ac-cording to Demanget et al. [30], adopting the following material con-stants: Young modulus E= 210000 MPa, Poisson's ratio ν = 0.3, YieldStress y = 1550 MPa. The superelastic behaviour of Nitinol wasmodeled using the corresponding built-in Abaqus user material sub-routine, while PET was modeled as an elastoplastic material. The ma-terial constants for these models are derived from the study of Klein-streuer and colleagues, who dealt with the computational mechanics ofNitinol stent-grafts [31]. A finite element mesh was created for each ofthe stent-grafts according to Auricchio et al. [24]; the details of eachmodel mesh are reported in Table 1.

2.3. Endograft deployment

The virtual deployment of the stent-grafts was performed within theaortic model derived from the medical image analysis, which wasconsidered as a rigid body. The rationale behind this choice is thatprevious studies demonstrated that the impact of wall elasticity isnegligible in the results of the deployment and differences are within anacceptable threshold for surgical planning [19,23,24]. Of note, if thefocus of the work was to estimate wall stresses and contact pressures, amore complete model of the vessel wall would be required [20]. An in-house Python code was used to create the input file for the finite ele-ment solver Abaqus/Explicit (Simulia, Dassault Systemes, France).

Firstly, the surface of the vessel is loaded into the workspace andsubsequently decimated (subsampled) so as to reduce the number ofelements and minimize the contact pairs during the simulation. Thecenterline of the vessel, previously calculated with VMTK, is alsoloaded. The user then manually selects the proximal landing pointwithin the centerline and the aorta, resembling the desired landing zoneof the endograft (see left panel of Fig. 3).

Secondly, a catheter that drives the stent-graft movement is built asa straight cylinder with a constant radius. Its most proximal section iscentered at the landing point previously selected by the surgeon. Theinitial diameter and length are the same as the outer diameter and totallength of the device to be implanted, respectively. The catheter is firstoriented in the direction of the normal defined by the proximal landingpoint and the immediately adjacent point in the centerline (see Fig. 3,center). Once the catheter is created, the stent-graft to be deployed isimported from the library and placed inside the catheter. Afterwards, africtionless contact pair is activated between the wires and the innersurface of the catheter.

Thirdly, the crimping phase takes place by simultaneously mod-ifying the orientation of the catheter and its radius. The center of theproximal-most section remains fixed to the landing point selected bythe user. The cylinder is then curved from its straight position to

Fig. 2. Computational representation of the proposed stent-grafts. A: CMD-TBE-40-70; B: CMD-ZTLP-A-40-65-1FF; C: ad-hoc designed device based on patient'sanatomy. All measurements are in mm.

Table 1Finite element domains of the three devices. M3D3 elements were used for themembrane and C3DR8 elements for the wires.

Stent-graft # of nodes # of membrane elements # of brick elements

A 96740 99280 28080B 118874 111118 37296C 156520 137060 51534

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

142

ultimately match the vessel centerline and, at the same time, the dia-meter is decreased to 13 mm resembling the stent-graft compressionwithin the delivery sheath. The displacement u uk k= is applied tothe catheter where

x y z x y zN

u P P( , , ) ( , , ) .kf i= (1)

In Equation (1), k represents the index of each time-step duringcrimping, x y zP ( , , )f is the final position based on the vessel centerline,

x y zP ( , , )i the initial position of each point of the cylinder and N thenumber of steps. Fig. 3 (center and right) illustrates the points of thecylinder from the initial to the crimped position.

The crimping stage of the analysis is performed in four time-stepswith a total duration of 0.03s, allowing to keep the ratio of the kineticand internal energy (ALLKE/ALLIE) under the threshold of 10%, en-suring a quasi-static regime of the analysis [22]. The rationale behindsplitting this simulation phase into four time-steps is based on an em-pirical observation aiming to reduce unphysical shape changes such aswrapping of the circular sections.

Finally, with the stent-graft in place, the catheter is expanded ra-dially to a diameter slightly wider than the native vessel. While doingso, a contact pair between the stent-graft struts and the luminal surfaceof the artery is activated. This procedure is performed in seven con-secutive time-steps with a total duration of 0.035s in order to avoidinertial effects. The computational time-step required for each iterationis automatically calculated by the software based on the materialproperties and the mesh characteristics. In order to reduce the com-putational time, mass scaling is activated to increase this time-step assuggested by the software vendor [32].

For the stent-grafts A and B, the catheter enlargement phase wasexecuted uniformly along the whole length of the device. This meansthat all the rings were radially expanded simultaneously. In the case ofdevice C, five different strategies were used for the expansion of thecatheter so as to account for the impact of the release sequence in thefinal positioning of the stent-graft. The deployment sequence is of ut-most importance since it conditions the overall fixation and thus par-ticular accuracy is needed, mostly due to the vicinity of the aortic valve(in the proximal side) and the supra-aortic vessels (in the distal end)[26]. Deployments were made: by uniformly expanding the catheter(uniform), with a distal to proximal expansion (backward), with aproximal to distal expansion (forward), starting by the central ring andpropagating the widening to the sides (center to sides) and finally re-producing a ring expansion resembling the multi-stage deployment ofcommercially available delivery systems (herein vendor). As an ex-ample, the intermediate steps followed by the catheter in the backward,center to sides and forward sequences are depicted in Fig. 4.

3. Results

Fig. 5 shows the final positioning of the three stent-grafts after thedeployment. In addition, the distance between the PET skirts and thevessel wall has been calculated as a measurement of the quality of theapposition. In devices A and B, higher separation with the wall is seenin the distal end of the stent-grafts.

All the consecutive steps of the enlargement of device C followed inthe vendor sequence are shown in Fig. 6.

To have a better understanding of the different enlargement se-quences that were used, Fig. 7 depicts the intermediate release step of

Fig. 3. Catheter orientation and movement during thecrimping phase. The blue sphere represents the proximallanding point selected by the user and is kept fixed duringall the procedure. The black arrow shows the normalvector used to initially place the cylinder representing thecatheter. Each (green) point within the cylinder followsthe trajectory given by the difference between the finalposition x y zP ( , , )f and its initial position x y zP ( , , )i , di-vided by the number of time-steps during this phase.

Fig. 4. Piecewise expansion of the various parts of the ca-theter following some of the deployment sequences. Top:backward deployment, starting from the arch and advancingtowards the annulus. Center: center to sides deployment,starting at the central ring of the device and expandingequally to the borders of the stent-graft. Bottom: forwarddeployment, starting from the annulus and propagating to-wards the distal ascending aorta (as used in most commondelivery systems).

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

143

all the deployment strategies.In order to account for the differences among release sequences,

Table 2 reports the average and maximum point-wise distances be-tween the position of the struts in device C. This choice was made sincethere was no baseline or gold standard to be used as a reference.

Finally, in order to ensure that all the simulations of device C wereperformed under a quasi-static regime, Fig. 8 depicts the ratio betweenthe kinetic energy (ALLKE) and internal energy (ALLIE) along the ex-pansion phase: uniform, forward and center to sides sequences ensured aratio under the 10% threshold along all the steps of the deployment.

4. Discussion

Ascending aortic pathologies are complicated to approach withTEVAR due to the anatomical complexity of this portion of the aorta.Even though many cases have been reported about successful outcomesof endografting procedures, as in the recent review of Muetterties andcolleagues [11], the technique is not yet consolidated. In this study weevaluated the computational deployment of both off-the-shelf and ad-hoc designed devices in order to predict the quality of stent-graft po-sitioning within a patient-specific scheme.

The biomechanics community has shown interest in endovascular

Fig. 5. Final positioning of the three stent-grafts within the ascending aorta. Distances between the devices skirts and the vessel are plotted in the last two rows usingdifferent views.

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

144

planning within virtual frameworks using patient-specific geometries,so as to predict surgery outcomes as well as choosing the right stent-graft. Contributions in the arch [23], abdominal [16,17,19,21] andthoracic aorta [22] have been analysed with different goals rangingfrom validating numerical frameworks and comparing devices by dif-ferent vendors to the development of very detailed computationalprocedures [20]. Given the less frequent interventions in the ascendingaorta only a previous contribution of our group has focused in thisparticular district [24]. Notwithstanding the variety of work done to thedate, ad-hoc realistic devices were not proposed by researchers, neitherwas a focus made in comparing deployment sequences.

From a more clinical point of view, during the course of this re-search, production of ascending aortic endografts by the companies wasfound to be limited to a short number of dedicated devices, mostly dueto the heterogeneity of geometries and lack of population studies thatguarantee stability of the prosthesis and treatment success. If results

prove to be concluding, computational modeling will encourage theproduction of off-the-shelf devices with more confidence.

From the results of the three endograft deployments, we can see thatdevice A lacks an appropriate apposition to the wall in the distal end.Even though the length of the device should be enough to divert flow inan aneurysm with analogue dimensions in the descending aorta, it failsto appropiately cover the aneurysmatic region due to its high angula-tion. A bird beak configuration is seen in the distal ring, where theseparation with the vessel wall is as high as mm7 . When compared todevices B and C, this behaviour could as well be attributed to the higherrigidity of Stainless Steel against Nitinol.

The second stent-graft, device B, was another model suggested by acompany to treat this case. The proximal ring was added in order toimprove stability in the proximity of the aortic valve leaflets withoutimpairing its opening. The choice of increasing the length of the cov-ered region by mm10 in comparison with device A has the goal of im-proving the distal apposition, however this was not completelyachieved.

Based on the results from the two previous devices, we designedprosthesis C by further extending the covered area and adding anotherdistal bare ring to improve fixation in the vicinity of the brachioce-phalic trunk without covering its ostium. In this case, the aneurysm wascompletely covered and both proximal and distal sealing improvedconsiderably. Special attention, however, has to be payed when usingbare rings in fragile aortas, since they have been associated with post-TEVAR retrograde dissection [33].

Fig. 6. Sequence representing a commercially available delivery system. Rings are not released in a sequential way but rather following a predefined pattern.

Fig. 7. Intermediate step of different deployment strategies in device C, all starting in the fully crimped state (Cr). Un: uniform, Ba: backward, Fo: forward, CS: centerto sides and Ve: vendor. Catheter was omitted to facilitate visualization.

Table 2Distances among the struts in the different deployment techniques, average/maximum in mm.

Backward Forward Center to sides Vendor

Uniform 1.10/4.9 0.36/2.6 0.37/1.8 0.95/3.2Backward NA 1.18/8.6 0.98/4.9 1.28/6.8Forward NA NA 0.36/2.7 0.92/4.6Center to sides NA NA NA 0.77/3.2

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

145

Another goal of this study was to prove whether different deploy-ment sequences would impact the final positioning of the stent-graft,challenged by the extreme angulation that makes deployment compli-cated. We ranged from a uniform expansion, which is computationallythe easiest and most intuitive, to sequential patterns of different com-plexity such as backward, forward and center to sides which coverfirstly the distal, proximal and central zones of the aneurysm respec-tively.

A sequence used in commercial devices was also evaluated to test itsimpact on the final apposition. Results from Table 2 show that meandifferences were at most above mm1 while considerable maximumdisplacements were mostly found the proximal side of the first tworings. Intuitively, forward and vendor strategies could be the easiest toapply in a clinical setting. To the best of our knowledge, the recent workby Berezowski and colleagues [26] is the first one in vascular surgeryliterature to raise a word of attention in the impact of the deploymentsequence, based on their in-vitro results. Even though the lack of a gold-standard makes us impossible to conclude which of the sequences is thebest one, the computational analysis showed that inertial effects aresignificant in many steps of the deployment. Therefore, when finiteelement simulations are performed with many mechanical variables,the impact in computational time should be considered.

5. Limitations

Some limitations of the work should be recognized when analysingthe results.

Firstly, a better standard for assessing the quality of the appositionshould be adopted since interactions between the aortic tissue and thestent-graft are a complex phenomenon. We currently base our conclu-sions on point-wise distance and disregard many of the wall mechanicalproperties. Therefore, to have a better appraisal of the results and asubsequent validation, the same deployments should be made in-vitrowithin a realistic aortic model as has been done in our previous work[24] and in literature focusing on abdominal aortic aneurysms[16,17,19]. These studies showed excellent agreement between thecomputational and the experimental data, confirming the robustnessand flexibility of FEA. This validation step should include differentaortas with a wide range of angulations, tortuosities, ages and gendersso as to confirm the applicability framework (for example, in Ref. [24]the deployment in the ascending aorta was performed in a female andin Ref. [23] virtual TEVAR was conducted both a male and a female). Inour case, due to the overall condition of the current patient, neither wasthe endovascular procedure carried out nor was device C manufactured,not allowing us to validate our results against experimental data.

A better assessment of the process could be done by using a de-formable model for the vessel wall. This would require the

characterization of aortic mechanical properties with particular atten-tion to diseased patients, who in general do not respond to classicalmaterial models.

No surgical tools, such as guide wires, other than the stent-graftitself have been considered in the present work. While the simplifiedintra-operative steps that we considered have only a minor influence onthe final device configuration, limitations for use in clinical practisehave to be acknowledged.

Stent-graft models were built as a unique solid geometry with themetallic rings and the graft as one piece. However, when devices areassembled by the manufacturer, the original diameter of the rings isreduced when sewed to the skirt, a phenomenon known as stress pre-deformation [29]. This configuration creates residual stresses in thestent-graft model that should be considered but were excluded in thiswork for simplicity. Furthermore, since we used a rigid-wall vessel, weomitted an analogous phenomenon occurring in the aorta when anelastic model is pressurized and a non-stress-free configuration isachieved. As a future work, stent pre-deformation can be included as asimulation step previous to our crimping sequence with the same ap-proach presented by Demanget and colleagues [30]. In this case, thestent and the graft start as separate solids where the initial phaseconsists on narrowing the Nitinol rings around the polymeric cylinder,from their manufactured width to the device nominal width. A contactpair is then activated between the skirt and the struts during shrinkingand finally the compression force is released, having both componentsattached. More complex pre-stressing strategies can be followed as inthe work of Hemmler et al. [20].

Long term stability of the devices is as important as the character-ization of intra-operative deployment. Only solid mechanics wereconsidered in this work, whereas a more comprehensive analysis in-cluding computational fluid dynamics would give better understandingon the displacement forces acting on the stent-graft [5].

6. Conclusion

Endografting in the ascending aorta is a challenging procedure,mostly for the associated comorbidities that patients subject to thisprocedure present. Appropriate selection of the device is crucial toensure a correct intra-surgical positioning and long-term stability. Inthis work we presented a case study of a patient who was unsuitable foropen repair of an ascending aorta pseudoaneurysm and therefore theonly choice was to undergo endovascular exclusion. Given the hostilebiomechanical environment of the vessel, we decided to test commer-cial devices as well as an ad-hoc developed stent-graft, which demon-strated a better apposition to the vessel wall. Comparison betweendeployment strategies used by different vendors demonstrated to havediverse computational behaviors, thus requiring special attention when

Fig. 8. Evolution of internal and kinetic energy during the deployment phase of device C. The total time of the deployment (Ttot) is divided into 7 enlargementphases in each sequence. A final uniform enlargement step was added to mitigate inertial effects.

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

146

large deformations within Finite Element Analysis are used. Finally, thisapplication of virtual endografting in the ascending aorta might en-courage companies to develop ad-hoc products for these diseases and,potentially, off-the-shelf endografts.

Conflicts of interest

Rodrigo M. Romarowski: None Declared.Michele Conti: None Declared.Simone Morganti: None Declared.Viviana Grassi: None Declared.Massimiliano M. Marrocco-Trischitta: None Declared.Santi Trimarchi: None Declared.Ferdinando Auricchio: None Declared.

Acknowledgements

This work was supported in part by ”Ricerca Corrente” and”5xmille” grants from IRCCS Policlinico San Donato, San DonatoMilanese, Milano, Italy.

References

[1] M. Grabenwöger, F. Alfonso, J. Bachet, R. Bonser, M. Czerny, H. Eggebrecht,A. Evangelista, R. Fattori, H. Jakob, L. Lönn, et al., Thoracic endovascular aorticrepair (tevar) for the treatment of aortic diseases: a position statement from theeuropean association for cardio-thoracic surgery (eacts) and the european society ofcardiology (esc), in collaboration with the european association of percutaneouscardiovascular interventions (eapci), Eur. J. Cardio. Thorac. Surg. 42 (1) (2012)17–24.

[2] M.M. Marrocco-Trischitta, G. Melissano, A. Kahlberg, G. Calori, F. Setacci,R. Chiesa, Chronic kidney disease classification stratifies mortality risk after electivestent graft repair of the thoracic aorta, J. Vasc. Surg. 49 (2) (2009) 296–301.

[3] C.A. Figueroa, C.A. Taylor, A.J. Chiou, V. Yeh, C.K. Zarins, Magnitude and directionof pulsatile displacement forces acting on thoracic aortic endografts, J. Endovasc.Ther. 16 (3) (2009) 350–358.

[4] S. Ishimaru, Endografting of the aortic arch, J. Endovasc. Ther. 11 (6_suppl) (2004)II–62.

[5] M.M. Marrocco-Trischitta, T.M. van Bakel, R.M. Romarowski, H.W. de Beaufort,M. Conti, J.A. van Herwaarden, F.L. Moll, F. Auricchio, S. Trimarchi, The modifiedarch landing areas nomenclature (malan) improves prediction of stent graft dis-placement forces: proof of concept by computational fluid dynamics modelling, Eur.J. Vasc. Endovasc. Surg. 55 (4) (2018) 584–592.

[6] M.M. Marrocco-Trischitta, H.W. de Beaufort, F. Secchi, T.M. van Bakel, M. Ranucci,J.A. Van Herwaarden, F.L. Moll, S. Trimarchi, A geometric reappraisal of proximallanding zones for thoracic endovascular aortic repair according to aortic arch types,J. Vasc. Surg. 65 (6) (2017) 1584–1590.

[7] G. Melissano, E. Civilini, L. Bertoglio, F. Calliari, F. Setacci, G. Calori, R. Chiesa,Results of endografting of the aortic arch in different landing zones, Eur. J. Vasc.Endovasc. Surg. 33 (5) (2007) 561–566.

[8] G. Antoniou, K. El Sakka, M. Hamady, J. Wolfe, Hybrid treatment of complex aorticarch disease with supra-aortic debranching and endovascular stent graft repair, Eur.J. Vasc. Endovasc. Surg. 39 (6) (2010) 683–690.

[9] T.M. van Bakel, H.W. de Beaufort, S. Trimarchi, M.M. Marrocco-Trischitta,J. Bismuth, F.L. Moll, H.J. Patel, J.A. van Herwaarden, Status of branched en-dovascular aortic arch repair, Ann. Cardiothorac. Surg. 7 (3) (2018) 406–413.

[10] R.R. Kolvenbach, A. Rabin, R. Karmeli, A. Alpaslan, E. Schwierz, Developments inparallel grafts for aortic arch lesions, J. Cardiovasc. Surg. 57 (3) (2016) 448–456.

[11] C.E. Muetterties, R. Menon, G.H. Wheatley, A systematic review of primary en-dovascular repair of the ascending aorta, J. Vasc. Surg. 67 (1) (2018) 332–342.

[12] R.C. Bernardes, T.P. Navarro, F.R. Reis, L.C. Lima, E.L. Monteiro, R.J. Procopio,F.E. Botelho, A. Dardik, Early experience with off-the-shelf endografts using a zone

0 proximal landing site to treat the ascending aorta and arch, J. Thorac. Cardiovasc.Surg. 148 (1) (2014) 105–112.

[13] E.E. Roselli, J. Idrees, R.K. Greenberg, D.R. Johnston, B.W. Lytle, Endovascularstent grafting for ascending aorta repair in high-risk patients, J. Thorac. Cardiovasc.Surg. 149 (1) (2015) 144–154.

[14] F.F. Mussa, S.A. LeMaire, J. Bozinovski, J.S. Coselli, An entirely endovascular ap-proach to the repair of an ascending aortic pseudoaneurysm, J. Thorac. Cardiovasc.Surg. 133 (2) (2007) 562–563.

[15] H. Krankenberg, R. Bader, S. Sixt, T. Tübler, C. Schwencke, D. Kivelitz, A. Rad,M. Caspary, T. Zeller, Endovascular repair of ascending aortic aneurysm by trans-apical approach and periscope technique, J. Endovasc. Ther. 20 (1) (2013) 13–17.

[16] S. De Bock, F. Iannaccone, G. De Santis, M. De Beule, D. Van Loo, D. Devos,F. Vermassen, P. Segers, B. Verhegghe, Virtual evaluation of stent graft deployment:a validated modeling and simulation study, J. Mech. Behav. Biomed. Mater. 13(2012) 129–139.

[17] D. Perrin, P. Badel, L. Orgéas, C. Geindreau, A. Dumenil, J.-N. Albertini, S. Avril,Patient-specific numerical simulation of stent-graft deployment: validation on threeclinical cases, J. Biomech. 48 (10) (2015) 1868–1875.

[18] D. Perrin, N. Demanget, P. Badel, S. Avril, L. Orgéas, C. Geindreau, J.-N. Albertini,Deployment of stent grafts in curved aneurysmal arteries: toward a predictive nu-merical tool, Int. J. Numer. Meth. Biomed. Eng. 31 (1) (2015) e02698.

[19] D. Perrin, P. Badel, L. Orgeas, C. Geindreau, S. rolland du Roscoat, J.-N. Albertini,S. Avril, Patient-specific simulation of endovascular repair surgery with tortuousaneurysms requiring flexible stent-grafts, J. Mech. Behav. Biomed. Mater. 63 (2016)86–99.

[20] A. Hemmler, B. Lutz, C. Reeps, G. Kalender, M.W. Gee, A methodology for in silicoendovascular repair of abdominal aortic aneurysms, Biomechanics Model.Mechanobiol. 17 (4) (2018) 1139–1164.

[21] R.M. Sanford, S.A. Crawford, H. Genis, M.G. Doyle, T.L. Forbes, C.H. Amon,Predicting rotation in fenestrated endovascular aneurysm repair using finite ele-ment analysis, J. Biomech. Eng. 140 (9) (2018) 091004.

[22] H.-E. Altnji, B. Bou-Saïd, H. Walter-Le Berre, Morphological and stent design riskfactors to prevent migration phenomena for a thoracic aneurysm: a numericalanalysis, Med. Eng. Phys. 37 (1) (2015) 23–33.

[23] R. Romarowski, E. Faggiano, M. Conti, A. Reali, S. Morganti, F. Auricchio, A novelcomputational framework to predict patient-specific hemodynamics after tevar:integration of structural and fluid-dynamics analysis by image elaboration, Comput.Fluids (2018), https://doi.org/10.1016/j.compfluid.2018.06.002 in press.

[24] F. Auricchio, M. Conti, S. Marconi, A. Reali, J.L. Tolenaar, S. Trimarchi, Patient-specific aortic endografting simulation: from diagnosis to prediction, Comput. Biol.Med. 43 (4) (2013) 386–394.

[25] D.E. Kendrick, A.F. Gosling, A. Nagavalli, V.S. Kashyap, J.C. Wang, Endovascularsimulation leads to efficiency and competence in thoracic endovascular aortic re-pair procedures, J. Surg. Educ. 72 (6) (2015) 1158–1164.

[26] M. Berezowski, S. Kondov, F. Beyersdorf, M. Jasinski, T. Plonek, M. Siepe,M. Czerny, B. Rylski, In vitro evaluation of aortic stent graft deployment accuracy inthe distal landing zone, Eur. J. Vasc. Endovasc. Surg. (2018), https://doi.org/10.1016/j.ejvs.2018.07.034 in press.

[27] M. Berezowski, J. Morlock, F. Beyersdorf, M. Jasinski, T. Plonek, M. Siepe,M. Czerny, B. Rylski, Inaccurate aortic stent graft deployment in the distal landingzone: incidence, reasons and consequences, Eur. J. Cardio. Thorac. Surg. 53 (6)(2017) 1158–1164.

[28] L. Antiga, M. Piccinelli, L. Botti, B. Ene-Iordache, A. Remuzzi, D.A. Steinman, Animage-based modeling framework for patient-specific computational hemody-namics, Med. Biol. Eng. Comput. 46 (11) (2008) 1097–1112.

[29] D. Roy, S. Lerouge, K. Inaekyan, C. Kauffmann, R. Mongrain, G. Soulez,Experimental validation of more realistic computer models for stent-graft repair ofabdominal aortic aneurysms, including pre-load assessment, Int. J. Numer. Meth.Biomed. Eng. 32 (12) (2016).

[30] N. Demanget, A. Duprey, P. Badel, L. Orgéas, S. Avril, C. Geindreau, J.-N. Albertini,J.-P. Favre, Finite element analysis of the mechanical performances of 8 marketedaortic stent-grafts, J. Endovasc. Ther. 20 (4) (2013) 523–535.

[31] C. Kleinstreuer, Z. Li, C. Basciano, S. Seelecke, M.A. Farber, Computational me-chanics of nitinol stent grafts, J. Biomech. 41 (11) (2008) 2370–2378.

[32] D. Systèmes, Abaqus Analysis Users Manual, Simulia Corp., Providence, RI, USA,2007.

[33] Y. Chen, S. Zhang, L. Liu, Q. Lu, T. Zhang, Z. Jing, Retrograde type a aortic dis-section after thoracic endovascular aortic repair: a systematic review and meta-analysis, J. Am. Heart Assoc. 6 (9) (2017) e004649.

R.M. Romarowski et al. Computers in Biology and Medicine 103 (2018) 140–147

147