Journal of Biomechanics

Transcription

1 Journal of Biomechanics 46 (2013) 7 12 Contents lists available at SciVerse ScienceDirect Journal of Biomechanics journal homepage: Realistic virtual intracranial stenting and computational fluid dynamics for treatment analysis Gábor Janiga a,n, Christian Rössl b, Martin Skalej c, Dominique Thévenin a a Laboratory of Fluid Dynamics and Technical Flows, University of Magdeburg Otto von Guericke, Germany b Visual Computing, University of Magdeburg Otto von Guericke, Germany c Institute for Neuro-Radiology, University of Magdeburg Otto von Guericke, Germany article info Article history: Accepted 31 August 2012 Keywords: Cerebral aneurysms CFD Hemodynamics Mesh deformation Virtual stenting VISC abstract In order to support the decisions of medical experts and to develop better stent designs, the availability of a simulation tool for virtual stenting would be extremely useful. An innovative virtual stenting technique is described in this work, which is directly applicable for complex patient-specific geometries. A basilar tip aneurysm provided for the Virtual Intracranial Stenting Challenge 2010 is considered to demonstrate the advantages of this approach. A free-form deformation is introduced for a wall-tight stent deployment. Numerical flow simulations on sufficiently fine computational meshes are performed for different configurations in order to characterize the inflow rate into the aneurysm and the corresponding residence time in the aneurysm sac. A Neuroform and a SILK stent have been deployed at various locations and the computed residence times have been evaluated and compared, demonstrating the advantage associated with a lower stent porosity. It has been found that the SILK stent leads to a large increase in the residence time and to a significant reduction in the maximum wall shear stress in the aneurysm sac. This is only observed when placing the stent in the appropriate position, showing that virtual stenting might be employed for operation support. & 2012 Elsevier Ltd. All rights reserved. 1. Introduction 1.2. Medical background 1.1. The Virtual Intracranial Stenting Challenge Each year, several groups of engineers associated with medical practitioners all around the world participate in a challenge relying only on computational techniques to assess the effect of stents for the treatment of cerebral aneurysms, see, e.g., Radaelli et al. (2008) or Cavazzuti et al. (2010). The main goal is to improve modeling techniques. Ultimately, the purpose is to provide valuable information to clinicians, so that both hemodynamic and anatomical factors could be considered before planning treatment. Under the auspices of the LINC and Seventh Intracranial Stenting (ICS10) Meeting, the Fourth Virtual Intracranial Stenting Challenge (VISC10) 1 was held in Houston, TX, September 13 16, Our research team from the University of Magdeburg Otto von Guericke in Germany has contributed to this challenge. n Corresponding author. Tel.: þ ; fax: þ address: janiga@ovgu.de (G. Janiga). 1 At present, a neurosurgeon must rely only on his/her own experience when deciding how to chooseandwheretoplaceastent to obtain an optimal treatment, i.e., flow alteration and consecutive thrombosis of the aneurysmal lumen. Even if this procedure is obviously successful, the support of a simulation tool during this critical process might help for untypical configurations and may serve as a mechanism facilitating and supporting the expert decision. Furthermore, an accurate simulation of blood flow before and after aneurysm treatment is necessary as a step towards the development of better and/or alternative stent designs. The successful treatment of an aneurysm by an implant should change the hemodynamics in the aneurysm sac, producing thrombogenic conditions, i.e., reducing the flow velocity and elongating the stasis. Numerical flow simulations based on computational fluid dynamics (CFD) may provide all important hemodynamic quantities. A first, qualitative examination can be performed by plotting selected contours of the shear stress, pressure or vector fields of the blood velocity. However, a quantitative analysis is essential in order to accurately quantify the effect of stent deployment and to select between possible alternatives. For this purpose, the flow stasis can be computed by considering the residence time within the sac, as it has been done by Seshadhri et al. (2011) for idealized and in Kim et al. (2008) for /$ - see front matter & 2012 Elsevier Ltd. All rights reserved.

2 8 G. Janiga et al. / Journal of Biomechanics 46 (2013) 7 12 both idealized and patient-specific geometries. Stent deployment should decrease the blood flow entering the aneurysm by excluding it from the arterial circulatory system. Several previous studies (Burleson and Turitto, 1996; Lanzino et al., 1999; Liepsch, 1986; Wakhloo et al., 1998) have already demonstrated that increasing the aneurysmal flow residence time can produce thrombus formation in cerebral aneurysms. Hence, to stimulate aneurysmal thrombosis, the increase in the stasis in the aneurysm must be targeted (Liou and Liou, 2004). This is the reason why residence time is used as a major indicator of stasis in the present work Patient configuration with a basilar tip aneurysm The organizers of the VISC10 challenge have provided to all participants the raw data of a basilar tip aneurysm found in a real patient. The considered geometry with the saccular aneurysm and the anatomy of the adjacent vessel tree is depicted in Fig. 1 after an accurate three-dimensional reconstruction by our group. In a further step, the quality of the surface mesh has been improved using an advancing front remeshing algorithm (Schöberl, 1997) in order to facilitate later generation of a high-quality volumetric mesh, while preserving all geometrical details of the vascular system. The aneurysm develops from the distal part of the basilar artery, the so-called basilar tip. The ostium is relatively wide compared to the dome and comprises the proximal parts (P1-segments) of both posterior cerebral arteries, especially that of the right side. Due to this geometry, an endovascular treatment with coiling alone would be difficult, if not impossible, without occlusion of the right P1- segment including the perforating arteries, that arise from this part of the posterior cerebral artery. Endovascular coiling is safe only after placement of a stent. On the other hand, aneurysm occlusion might also be achieved by deployment of a flow diverter without coiling, as will be considered in what follows. 2. Materials and methods The main motivation of this study is to investigate in a quantitative and realistic manner the effect of stent deployment in real, patient-specific geometries. In this manner, it should ultimately become possible to prevent the rupture of intracranial aneurysms by supporting in the best possible way the decision of medical experts. At the same time, a numerical tool for virtual stenting would be useful to guide the development and optimization of new implant designs. In this section, all methods needed to solve such a problem are described and used considering the configuration of the VISC10 challenge. The image segmentation was first performed using the automatic pipeline developed by our group (Moench et al., 2011). The computational geometry has been smoothed while keeping all important features and exported as an STL surface mesh. The stent geometries are reproduced using the CAD software Autodesk Inventor (Autodesk, San Rafael, USA). A SILK (Balt International, Montmorency, France) flow diverter and a Neuroform (Boston Scientific Neurovascular, Fremont, CA) stent were placed in two different positions (AD and AE vessel branching in Fig. 1). The resulting deployment, as described later, comes in direct wall-tight contact with the vessel walls in spite of the resulting, complex deformation. The porosity of the applied stents was evaluated integrating the area when unrolling the stent onto a plane. The Neuroform stent had a high porosity of 84% (i.e., a low strut density), while the SILK stent had a considerably lower porosity of 60% (higher strut density). Further comparisons of these implants can be found for instance in Seshadhri et al. (2011) Stent deployment Virtual stent deployment has been already considered successfully in the nineties (Aenis et al., 1997) for simplified configurations (usually, a straight pipe). The objective of the present paper is to develop a virtual stenting technique applicable for real patient geometries. The goal of virtual stent deployment is to first find a rigid transformation corresponding to stent extension followed by a non-rigid deformation, aligning the deployed stent geometry by following the centerline of the vessel geometry. This deformation leads to a tight fit of the stent so that it is in direct contact with the vessel walls. Prior publications (Appanaboyina et al., 2009; Larrabide et al., 2012) rely on a deformable shape model (see, e.g., McInerney and Terzopoulos, 1993, 1996) used to deform a cylindrical support surface. An alternative approach (Gundert et al., 2011) represents the stent as a solid model and applies constructive solid geometry (CSG) operations for deployment. Augsburger et al. (2011) proposed a different technique based on the modeling of the flow diverter as a porous medium. A recent approach (Bernardini et al., 2011, 2012) simulates deployment based on a finite element model of the stent and the vessel. In contrast to this, the approach proposed here applies a non-rigid registration based on a free-form deformation. According to the authors knowledge, such a free-form mesh deformation technique has never been applied for virtual stenting yet. This new approach is conceptually easier, involves less free parameters, and leads to physically plausible results as well. The main objective is to speed up virtual stent deployment while reproducing the true conditions of operation planning. A fast process is important because the expertise of a medical specialist is required, who is typically available only for a short period of time. Therefore, the computational tools should show interactive response, if possible in real-time. Obviously, the goal of a highly efficient virtual deployment is in conflict with the goal of having a simulation as realistic as possible. However, a truly exact description of stent deployment would only be possible when knowing in detail all material properties of the vessel walls. Unfortunately, there is at present no method available to measure these properties for practical cases. Hence, a phenomenological approach appears to be the best possible solution. The presented approach finds in a very short time a configuration, which corresponds to the experience of medical practitioners, guided by years of corresponding interventions. The deployment proceeds in two steps. First, a rough, large scale deformation makes the stent follow the smoothed centerline curve extracted from the vessel data. Second, this initial deformation is refined iteratively such that the stent geometry gets closer and closer to the real vessel boundaries. Both steps rely on free-form deformations (Sederberg and Parry, 1986) based on trivariate triharmonic splines, similar to the thin-plate splines (Bookstein, 1989; Duchon, 1977) employed in 2D. For determining the parameters of the deformation the stent model is temporarily replaced by a cylindrical geometry that incorporates the convex hull of the stent before deployment. The resulting deformation is finally applied by replacing the cylindrical template with the real stent model. A similar approach was taken by Appanaboyina et al. (2009), whereas Larrabide et al. (2012) assume a set of rings as a template. An innovative virtual stent deployment using a free-form deformation is introduced in this work (Supplementary data). Further details of this method are described in Botsch and Kobbelt (2005) Computational mesh The wall-tight stent deployment is exemplified in Fig. 2 for a SILK flow diverter. The SILK stent is composed of a dense strut network with 48 wires (8 wires have a diameter of 50 mm and 40 have a diameter of 30 mm). The obtained virtual stent deformation is illustrated considering both outside and inside views. All computational meshes needed for the later finite volume simulations have been generated using the commercial tool ANSYS IcemCFD (Ansys Inc., Canonsburg, Fig. 1. (a,b) The anatomical model of the computational configuration after reconstruction. (c) Original 2D DSA image of the basilar tip aneurysm provided by the organizers of VISC10.

3 G. Janiga et al. / Journal of Biomechanics 46 (2013) Fig. 2. Wall-tight stent deployment using a SILK flow diverter. (a) Deployment following the AE vessel (view from outside), (b) inside view from the artery and (c) inside view from the aneurysm. 3. Results Fig. 3. Computational mesh in the virtually stented domain. PA, USA). The obtained grid quality has been carefully checked and maintained within the optimal range, determined from previous studies concerning mesh sensitivity and mesh independence (Janiga et al., 2009). The generated body-fitting meshes involve up to 13,500,000 finite volume cells composed of tetrahedral elements in order to reduce the discretization errors. An extremely fine resolution is obtained in the vicinity of the stent struts, so that their geometry is appropriately reproduced by the numerical mesh. The mesh generation for the struts of a flow diverter stent is challenging and it is only feasible if the surrounding region is already finely discretized in order to maintain grid quality. The volume of the smallest grid volume element corresponds to 26:5 mm 3. Fig. 3 shows the final volume mesh generated after stent deployment. The very high resolution and quality of this mesh is exemplified by an enlarged cut through the aneurysm sac (Fig. 3, right). The finer mesh cells mark the stent location Computational details All fluid flow simulations have been performed using the commercial software ANSYS Fluent 12 (Ansys Inc., Canonsburg, PA, USA). Only steady state computations are reported here, the analysis of the performed time-dependent simulations is not considered for the sake of brevity. While the real flow is obviously unsteady, further studies by our group and other publications (see, e.g., Kim et al., 2008) indicate that this assumption is acceptable to compare the stenting efficiency of different implants. A purely laminar flow simulation is employed because the Reynolds number at the inlet is 470 based on the mean velocity and hydraulic diameter. At the inlet of the computational domain a fully developed velocity profile (corresponding to an infinitely long pipe with a uniform cross-section) is given as a Dirichlet boundary condition (prescribed velocity values). The fixed inlet flow rate amounts to 2:36 ml s 1, corresponding to the value provided for VISC 09. Traction-free conditions have been taken into account at all the outlets, assuming an identical uniform relative pressure. Since no further details are known concerning the further course of the vessels, this assumption is appropriate. A standard, no-slip boundary condition is employed at all contact points with surfaces (vessels, stent) and all vascular walls are assumed to be rigid. Blood rheology is represented using a Newtonian description with constant density and viscosity, where the blood density is chosen as 1:05 g=cm 3 and the dynamic viscosity as 4 mpas. Previous studies by our group and independent publications (see, e.g., Fisher and Rossmann, 2009) indicate that the aneurysm morphology has a much larger influence on the outcome than the non-newtonian behavior for such relatively large vessels. Computations are carried out in parallel using eight Linux computing cores (2.1 GHz AMD Opteron 64-bit dual-quad processors). For the finest mesh considered in this study, 19.5 GB of computer memory are needed for the simulation. Second-order accurate, fully converged results have been obtained in about 20 h of computing time, reaching normalized residuals of At present, there is still no agreement in the scientific community concerning a unique, appropriate quantity that could be employed to predict the rupture risk of an aneurysm (Cebral et al., 2011a, 2011b; Xiang et al., 2011). Since the emphasis is clearly set here on the treatment options and hence on thrombosis, the flow rate exchanged through the ostium between the main vessels and the aneurysm sac has been retained as a major indicator of treatment quality. This flow rate is evaluated in a postprocessing step using EnSight (CEI Inc., Apex, USA) integrating the velocity field along the same plane (associated to the ostium) for all the considered configurations. A successful treatment should reduce as much as possible this quantity and consequently increase the residence time within the aneurysm, leading to optimal thrombotic conditions. Selected streamlines of the computed flow are shown in Fig. 4 for all the considered cases, demonstrating the qualitative reduction in the flow entering the aneurysm sac after stenting, in particular when following the AD vessel. The corresponding volume flow rates entering through the neck of the aneurysm are compared in Table Discussion An innovative virtual stenting technique is introduced in this work, which is directly applicable for complex patient-specific geometries. In order to validate this deployment technique, the considered flow-diverter (SILK) has been characterized experimentally after deployment in an in-house, silicone phantom model based on a real patient geometry. Two different experimental techniques, DynaCT (Biplanar C-Arm System, Axiom Artis, Siemens, Forchheim, Germany) and Micro-CT (Phoenix Nanotom S, GE Sensing & Inspection Technologies GmbH, Wunstorf, Germany) have been tested for this purpose. Even if a full resolution of the finest struts is not possible due to the limited spatial resolution, the position and overall geometry of the deployed stent is extremely well resolved, allowing a direct comparison with the virtual deployment procedure. The geometries obtained experimentally have been reconstructed and registered into a common coordinate system, corresponding to the numerical framework. Fig. 5 shows the results of the virtual and real deployments in the same phantom model. Comparing the figures an excellent agreement can be observed. Hence, the introduced virtual stent deployment technique leads to a very faithful representation of the true deployment process. The application of this method is now presented for the basilar tip aneurysm retained for VISC10 and the resulting hemodynamic modifications are identified. The flow stasis within the aneurysm has been computed from the residence time (Kim et al., 2008; Seshadhri et al., 2011) and is also listed in Table 1. The associated residence time T is determined by dividing the full aneurysm volume V by the inlet volume flow rate _q in found at the aneurysm

4 10 G. Janiga et al. / Journal of Biomechanics 46 (2013) 7 12 Fig. 4. Selected streamlines (a) without and with (b, c) Neuroform and (d, e) SILK stents: (a) Streamlines without stent. (Inflow rate: 0.99 cm3/s), (b) neuroform in AE vessel. (Inflow rate: 0.93 cm3/s), (c) neuroform in AD vessel. (Inflow rate: 0.80 cm3/s), (d) SILK in AE vessel. (Inflow rate: 0.84 cm3/s) and (e) SILK in AD vessel. (Inflow rate: 0.56 cm3/s). Table 1 Comparison of the inflow rates and the residence times for the different cases. Stent Figure number Inflow rate (cm 3 /s) Residence time (s) Ratio of residence time Without stent 4(a) Reference (1.00) Neuroform AE 4(b) Neuroform AD 4(c) SILK AE 4(d) SILK AD 4(e) neck, defined as a straight plane placed slightly above the stent, at the level of the ostium: T ¼ V= _q in. The aneurysmal inflow rate is obtained by integrating the entering flow rate over this cross section. The computational results have been analyzed and compared with the reference case, before stent deployment. Though the Neuroform stent is not considered to act as a flow diverter due to its high porosity, a noticeable increase in the residence time (25%) has been observed when stenting with Neuroform through the AD vessel. On the other hand, the same stent deployed in the AE vessel does not lead to a noticeable improvement. As expected, the SILK flow diverter succeeds best in reducing the inflow rate within the aneurysm sac, i.e., in increasing the residence time. Similar to the Neuroform stent, the impact of SILK placed in the AE vessel is considerably reduced, and comparable to that of the Neuroform stent deployed in the AD vessel. Stenting through the AD vessel obviously appears to be the right choice. The best solution is finally the deployment of a SILK flow diverter within the AD vessel. In this way, the residence time within the aneurysm sac is increased by 77%, a very high value. The above results demonstrate that although a flow diverter stent is expected to be more efficient when trying to induce thrombosis in the aneurysm, an appropriate position is essential as well. The AD vessel must be selected for deploying the implant in the considered case. Obviously, the developed methodology will never replace the experience of medical specialists. The medical partners of this project identified the best strategy (deploying a SILK flow diverter following the AD vessel) within minutes. This was finally found as well in the presented numerical simulations, but after some weeks of work. Nevertheless, this agreement confirms that the

5 G. Janiga et al. / Journal of Biomechanics 46 (2013) Fig. 5. Flow-diverter stent (SILK) in an in-house, silicone phantom model based on a real patient geometry. (a) Virtual stenting based on our free-mesh deformation, (b) deployed SILK stent measured experimentally by DynaCT, (c) deployed SILK stent measured experimentally by Micro-CT and (d) all views together (using same color code). simulation tools developed for this study lead to a correct representation of hemodynamic changes induced by stenting and might therefore be very useful: to support decisions of medical experts for very unusual patient geometries, or when a variety of stent and flow diverter models will become available on the market; to help develop new implants, even better suited for inducing thrombosis in cerebral aneurysms (perhaps patient-specific in a few years?); to identify new treatments for complex cases like giant aneurysms, multiple aneurysms or aneurysms located at vessel bifurcations. Furthermore, the available computing power is rapidly growing and parallelizationmightbeusedtoamuchlargerextentthaninthe present case. Overall, virtual stenting should ultimately become possible as decision support within hours for an elective treatment. Conflict of interest statement We declare that we have no conflict of interest. Acknowledgments The Land Saxony-Anhalt in Germany has financially supported this work through the collaborative project MoBeStAn. The authors thank M. Neugebauer and R. Gasteiger for reconstruction of the geometry and Dr. O. Beuing for interesting discussions. The help of S. Seshadhri to produce the CAD geometries of the stents and the Micro-CT experiment provided by S. Rannabauer is highly acknowledged. Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at References Aenis, M., Stancampiano, A.P., Wakhloo, A.K., Lieber, B.B., Modeling of flow in a straight stented and nonstented side wall aneurysm model. Journal of Biomechanical Engineering 119, Appanaboyina, S., Mut, F., Löhner, R., Putman, C., Cebral, J., Simulation of intracranial aneurysm stenting: techniques and challenges. Computer Methods in Applied Mechanics and Engineering 198, Augsburger, L., Reymond, P., Rufenacht, D., Stergiopulos, N., Intracranial stents being modeled as a porous medium: flow simulation in stented cerebral aneurysms. Annals of Biomedical Engineering 39, Bernardini, A., Larrabide, I., Morales, H.G., Pennati, G., Petrini, L., Cito, S., Frangi, A.F., Influence of different computational approaches for stent deployment on cerebral aneurysm haemodynamics. Interface Focus 1, Bernardini, A., Larrabide, I., Petrini, L., Pennati, G., Flore, E., Kim, M., Frangi, A., Deployment of self-expandable stents in aneurysmatic cerebral vessels: comparison of different computational approaches for interventional planning. Computer Methods in Biomechanics and Biomedical Engineering 15, Bookstein, F., Principal warps: thin-plate splines and the decomposition of deformations. IEEE Transactions of Pattern Analysis and Machine Intelligence 11, Botsch, M., Kobbelt, L., Real-time shape editing using radial basis functions. Computer Graphics Forum 24, Burleson, A.C., Turitto, V.T., Identification of quantifiable hemodynamic factors in the assessment of cerebral aneurysm behavior. Thrombosis and Haemostasis 76,

6 12 G. Janiga et al. / Journal of Biomechanics 46 (2013) 7 12 Cavazzuti, M., Atherton, M., Collins, M., Barozzi, G., Beyond the Virtual Intracranial Stenting Challenge 2007: non-newtonian and flow pulsatility effects. Journal of Biomechanics 43, Cebral, J.R., Mut, F., Weir, J., Putman, C., 2011a. Quantitative characterization of the hemodynamic environment in ruptured and unruptured brain aneurysms. American Journal of Neuroradiology 32, Cebral, J.R., Mut, F., Weir, J., Putman, C.M., 2011b. Association of hemodynamic characteristics and cerebral aneurysm rupture. American Journal of Neuroradiology 32, Duchon, J., Spline minimizing rotation-invariant semi-norms in Sobolev spaces. In: Schempp, W., Zeller, K. (Eds.), Constructive Theory of Functions of Several Variables, pp Fisher, C., Rossmann, J.S., Effect of non-newtonian behavior on hemodynamics of cerebral aneurysms. Journal of Biomechanical Engineering 131, Gundert, T., Shadden, S., Williams, A., Koo, B.K., Feinstein, J., LaDisa, J., A rapid and computationally inexpensive method to virtually implant current and next-generation stents into subject-specific computational fluid dynamics models. Annals of Biomedical Engineering 39, Janiga, G., Beuing, O., Seshadhri, S., Neugebauer, M., Gasteiger, R., Preim, B., Rose, G., Skalej, M., Thévenin, D., Virtual stenting using real patient data. In: J. Vad (Ed.), Conference on Modelling Fluid Flow, Budapest, Hungary, pp Kim, M., Taulbee, D.B., Tremmel, M., Meng, H., Comparison of two stents in modifying cerebral aneurysm hemodynamics. Annals of Biomedical Engineering 36, Lanzino, G., Wakhloo, A.K., Fessler, R.D., Hartney, M.L., Guterman, L.R., Hopkins, L.N., Efficacy and current limitations of intravascular stents for intracranial internal carotid, vertebral, and basilar artery aneurysms. Journal of Neurosurgery 91, Larrabide, I., Kim, M., Augsburger, L., Villa-Uriol, M.C., Rüfenacht, D., Frangi, A.F., Fast virtual deployment of self-expandable stents: method and in vitro evaluation for intracranial aneurysmal stenting. Medical Image Analysis 16, Liepsch, D.W., Flow in tubes and arteries a comparison. Biorheology 23, Liou, T.M., Liou, S.N., Pulsatile flows in a lateral aneurysm anchored on a stented and curved parent vessel. Experimental Mechanics 44, McInerney, T., Terzopoulos, D., A finite element model for 3D shape reconstruction and nonrigid motion tracking. In: Proceedings of the Fourth International Conference on Computer Vision (ICCV 93), Berlin, Germany, pp McInerney, T., Terzopoulos, D., Deformable models in medical image analysis: a survey. Medical Image Analysis 1, Moench, T., Gasteiger, R., Janiga, G., Theisel, H., Preim, B., Context-aware mesh smoothing for biomedical applications. Computers & Graphics 35, Radaelli, A., Augsburger, L., Cebral, J., Ohta, M., Rüfenacht, D.A., Balossino, R., Benndorf, G., Hose, D.R., Marzo, A., Metcalfe, R., Mortier, P., Mut, P., Reymond, P., Socci, L., Verhegghe, B., Frangi, A.F., Reproducibility of haemodynamical simulations in a subject-specific stented aneurysm model a report on the virtual intracranial stenting challenge Journal of Biomechanics 41, Schöberl, J., NETGEN: An advancing front 2d/3d-mesh generator based on abstract rules. Computing and Visualization in Science 1, Sederberg, T.W., Parry, S., Free-form deformation of solid geometric models. In: Proceedings of the SIGGRAPH, Dallas, pp Seshadhri, S., Janiga, G., Skalej, M., Thévenin, D., Impact of stents and flow diverters on hemodynamics in idealized aneurysm models. Journal of Biomechanical Engineering 133, /1 9. Wakhloo, A.K., Lanzino, G., Lieber, B.B., Hopkins, L.N., Stents for intracranial aneurysms: the beginning of a new endovascular era? Neurosurgery 43, Xiang, J., Natarajan, S., Tremmel, M., Ma, D., Mocco, J., Hopkins, L., Siddiqui, A., Levy, E., Meng, H., Hemodynamic morphologic discriminants for intracranial aneurysm rupture. Stroke 42,