A numerical analysis of a semi-enclosed tubular mechanical embolus retrieval device (MERD) for the treatment of acute ischemic stroke (AIS) is presented. In this research, the finite element analysis (FEA) methodology is used to evaluate mechanical performance and provide suggestions for optimizing the geometric design. A MERD fabricated from nickel–titanium alloy (Nitinol) tubing is simulated and analyzed under complex in vivo loading conditions involving shape-setting, crimping, deployment, and embolus retrieval. As a result, the peak strain of the shape-setting procedure is proved to be safe for the device pattern. However, the MERD shows poor mechanical behavior after crimping into a catheter, because the peak crimping strain obtains a value of 12.1%. The delivery and deployment step demonstrates that the artery wall has little risk of serious injuries or rupture. In addition, the process of simulation of embolus retrieval and device system migration inside the cerebral artery lumen provides useful information during the design process.
Stroke is a type of brain function disease which is due to the disturbance of blood supply to the brain tissue. Stroke has already become the second leading cause of death in the world. Of yearly stroke occurrences, approximately 88% are acute ischemic stroke (AIS) . It is reported that the number of deep venous thrombosis cases has increased to 250,000 per year in the U.S., and the ratio of morbidity or mortality is also fairly high . In order to minimize nervous system damage, a successful treatment is expected to remove plaque emboli and reopen the stenosed blood vessel so as to restore blood flow before irreversible damage occurs . The desirable therapeutic method for AIS is chosen on the basis of the therapeutic time window. The therapeutic time window of a standalone mechanical thrombectomy has to be within 6 h in the treatment of AIS . If so, AIS-induced death and serious injury can be reduced. Mechanical embolus retrieval devices (MERDs) are being extensively used to treat stroke in clinics. Thrombectomy has proven to reduce the risk of operation mishaps as compared to conventional open surgery, and this strategy provides a less invasive therapy in the treatment of AIS. MERD structures vary in complexity from a three-dimensional (3D) helical coil to a laser-cut nickel–titanium (Nitinol) tube. So far, several embolus retrieval devices have been used in clinical applications, such as the Merci retrieval system, the Solitaire FR, and others [5,6]. The multilattice tubular structure of these devices is often manufactured from thin-walled Nitinol tubing by laser-cutting. The carved meshlike pattern is then electropolished to remove burrs and eliminate excess material to meet the size allowance. In order to build a semiclosed metal tube, the four distal forks are constrained and connected together. During heat-treatment by annealing, the Nitinol device's final shape and size are achieved. The proximal strut is connected with the pushing guide wire. After the prescribed process operations, the MERD is shrunk into a catheter and delivered to the stenosed cerebral vessel. Once the MERD has been delivered to the desired arterial segment, the outer catheter is removed gradually. Meanwhile, the MERD self-expands to make contact with the surrounding vessel wall. The desired tubular mesh structure exerts a radial force on the artery wall to keep it open, while the colander-shaped strut captures the crushed plaque emboli, preventing those segments from escaping. Finally, the embolus retrieval system, along with the plaque emboli, is manipulated so that it migrates inside the path vessel. The superelasticity, shape-memory, biocompatibility, fatigue resistance, and corrosion resistance properties of this material have made Nitinol attractive for intracranial medical applications.
Some previous researchers have studied the mechanical behavior of Nitinol meshlike tubular devices. Krischek et al. report experimental outcomes comparing the physical properties of Nitinol intracranial stents. These functional features include the radial force, wall apposition, conformability, Gator backing effects (a stent's tendency to flair its struts outward, forming protrusions into convexity), kinking behavior, and ovalization . Azaouzi et al. developed a numerical analysis approach for the gradual deployment procedure of self-expanding stents, allowing smooth contact effects between the stent and the vessel wall . Wu et al. built a finite element model of stents and studied the axial flexibility by applying moment loading . Grogan et al. designed three types of unique CoCr stents and performed a physical comparison of the deployment, radial force, longitudinal resistance, and flexibility. Also, the device-related expansion procedure for a rigid cylinder by a polymer balloon is also reported . Kleinstreuer et al. simulated the crimping, deployment, and cyclical loading conditions of a self-expanding Nitinol stent. Most important of all, the strain-induced Nitinol high-cyclic fatigue life of stents due to vessel movement is estimated and analyzed with a mean strain/alternating strain diagram. Based on the blood vessel wall compliance, the rubberlike hyperelastic constitutive material model was established .
Systematic research on the mechanical performance of embolus retrieval devices, however, has not yet been published. After delivery and insertion into the blood vessel, Nitinol stents expand and counteract the effects associated with diseased arteries; that is, the tubular stent embeds itself in the stenosed artery wall segment to avoid distal movement. By contrast, the lattice-shaped MERD is manipulated so that it migrates inside the blood lumen to remove the plaque blockage. In the existing research on crimping and deployment, the mechanical behavior is simplified by applying displacement to the movable crimping cylinder. In this paper, the MERD pattern is compressed and expanded by the axial movement of a rigid cylindrical tool, which provides a more realistic and accurate numerical modeling. Poor performance of the embolus-capturing system migration may result in undesired clinical accidents and even plaque retrieval failure. Hence, the capability for system movement is particularly crucial during operation. Furthermore, the modeling of the process of embolus retrieval has not been reported thus far. Also, MERD works as a disposable surgical device for thrombectomy rather than as a permanent implant. Therefore, the predicted device life of the Nitinol MERD is fully dependent on the peak strain. Finally, a unique distal structure constructed could be considered for MERD designs. Modeling of tubular geometry and the embolus retrieval procedure as well as a failure-evaluation strategy could allow a systematic scientific numerical workflow analysis that can help optimize the tube structure.
A simplified finite element analysis (FEA) model (half of the axisymmetric MERD) is used to simulate shape-setting, crimping, and self-expansion as well as the embolus retrieval process. The common adverse events are listed as follows: (1) cracking or fracturing of struts, (2) loss of deliverability, (3) serious trauma or rupture of vessel wall, and (4) loss of efficiency for plaque retrieval inside the cerebral lumen. The device lifetime of a Nitinol intracranial surgical instrument is evaluated and predicted on the basis of the maximum principal strain (MPS). A relatively high-magnitude strain distribution (common in “flaw-prone” regions) and peak MPS are obtained and utilized to predict the lifetime of a Nitinol MERD. The core numerical simulation conditions involve expansion and annealing for the shape-setting status, crimping or shrinking into the catheter, self-expansion inside the cerebral vessel, and embolus retrieval as well as device system migration. The workflow of MERD after delivery to the targeted artery is shown in Fig. 1.
Applying FEA to a Nitinol MERD provides both a device lifetime prediction workflow and a structure optimization scheme. A commercial FE code solver, abaqus 6.10/standard (DS SIMULIA, Johnston, RI), and the subroutine (UMAT/Nitinol) are employed to solve contact interaction equations. In Sec. 2, the MERD geometry and FEA model as well as the constitutive equations of the engineering materials are presented. The simulation results for in vivo loading are illustrated in Sec. 3. In this section, the structure failure prediction scheme and geometry optimization are demonstrated. Section 4 discusses the mechanical effects of a tubular MERD in the shape-setting, crimping, self-expansion, and embolus retrieval procedure. The limitations of the present research are also described.
Materials and Methods
The nickel–titanium alloy is becoming attractive in interventional medical applications. The beneficial characteristic of superelasticity is utilized to realize shape-setting and self-expansion at human body temperature. At 37 °C, Nitinol comprises the metal in austenite phase. Superelastic Nitinol shows endurance of large strain-dominated deformation in uploading condition and recovery to the original shape-size during unloading. When the maximum principal strain (MPS) quantitative value reaches 6%, Nitinol metal is located in the superelastic domain; that is, a complete shape-recovery can be achieved upon unloading. Nitinol's strain threshold value is considered to be 12%. An output strain field ranging from 6% to 12% is regarded as acceptable, ignoring fatigue fracture.
There are numerous embolus retrieval device designs, which vary in complexity in clinical use: some structures comprise a cylinder helical coil, while others are a multilattice structure for removing materials on Nitinol tubing [5,6]. In this paper, a new meshlike tubular embolus retrieval structure is discovered. Figures 2(a) and 2(b) show the planar computer aided design sketch with which the pattern is laser-carved. The parameters include strut width (SW), fillet diameter (FD), strut length (SL), and so on. Generally, the tubular mesh structure consists of several “rhombus” diamond cells. As is shown in Fig. 2(c), the device geometric model comprises five “columns” in the axial direction. Each “column” consists of four diamond cells in a circumferential row.
The desired blood vessel wall is built as a cylindrical rubbery tube with a thickness of 0.1 mm and an inner diameter of 3.0 mm, respectively. A plaque embolus is shaped like a bullet, combining a cylinder and a hemisphere. The shape design improves the continuity of the geometry and aids analysis convergence. To construct the semiclosed pattern, the FE model of colander-shaped struts is meshed and connected to the principal body after the shape-setting procedure.
In order to balance computation costs and analysis accuracy, the axial symmetrical model is built. The geometry model of the MERD is meshed by C3D8R (three-dimensional eight-node stress hex element with reduced integration) . C3D8R should reduce the computing costs as well as improve convergence. Hourglass control is applied to avoid undesired analysis effects. The meshing densities of three seeds along the strut thickness direction and eight seeds along the fillet curve are utilized to ensure the validity of the results. The element and node numbers of the MERD FE model are 40,000 and 20,000, respectively. To represent hyperplastic behavior, the arterial wall and the plaque embolus are modeled with the element type C3D8H (an eight-node linear brick which is hybrid with constant pressure). The shape-setting tool is assumed to be a semirigid progressive cylinder with the element type SFM3D4 (3D linear four-node bilinear rigid quadrilateral element). And the crimping catheter is modeled as an analytical rigid body. Before starting the simulation, the mesh density independence study is performed.
This research provides a computational simulation scheme to investigate the mechanical performance of a MERD under in vivo loadings.
Shape-setting is applied to generate a final-sized MERD pattern from original tubing, using the expansion tool in the radial direction. This original 2.00 mm outer diameter (OD) Nitinol tubing is expanded to the nominal dimension of 4.0 mm by a progressive movable cylinder. A master/slave contact pair is generated between the expansion tool exterior surface and the MERD interior surface. All nodes of the assembly are transformed under a global cylindrical coordinate system. One node in the middle segment is restrained in the axial direction. Meanwhile, the nodes of the middle symmetrical section are also restrained in a circumferential row. The degrees-of-freedom (DOF) of X axial, Y axial, and Z axial of the semirigid cylinder are limited to prevent transitional axial and rotational movement. An outward radial displacement imposed on the cylindrical punch can lead to tubular pattern expansion. Figure 3(a) shows the schematic view of MERD during shape-setting step.
When crimped into a catheter, the tubular pattern encounters significant deformation, which might result in failure of the structure. A global cylindrical coordinate system is selected. The axial displacement is applied to the proximal tip of the MERD. The axial symmetry section nodes are also constrained. The reference point of the rigid catheter is assumed to be fixed in all DOFs; that is, this limits the rigid body's motion in all directions. Figure 3(b) shows the schematic view of MERD during crimping step. The contact algorithm of the surface-to-surface pair utilizes master surface priority for the crimping punch, while the MERD outer surface is set as the slave surface. It is notable that self-contact should also be set to prevent struts overlapping.
After the MERD is crimped and delivered to the targeted artery, it self-expands and recovers to its nominal dimensions. In the simulation of this step, the present instances include a crimping catheter, the MERD, and the blood vessel wall. The embolus retrieval device is deployed and inserted inside the blood carrying lumen by removing the crimping catheter and eliminating the contact effect. The reference point of the catheter imposes an axial displacement. In the deployment process, the proximal tip of the device is also fixed axially for an accurate location. The symmetrical sections of the MERD and blood vessel are fixed to prevent rotation. Also, the proximal and distal sections of the tubular artery are also restrained in the axial and radial directions to prevent them from moving. When the MERD is separated progressively out of the catheter, it recoils and penetrates into the vessel wall. The artery experiences a supporting force due to mesh pattern shape-recovery. A contact algorithm provides master surface priority to the MERD exterior surface, while the arterial interior surface is designed as the slave surface. For the added contact pair, a friction coefficient of 0.25  is employed.
After delivery and insertion of the MERD inside the arterial wall, the bulletlike plaque embolus is added and activated inside the semiclosed tubular MERD. The colander-shaped struts capture and remove the plaque embolus when the MERD system is withdrawn. In this mode, the newly added contact pairs are listed as follows: (1) the embolus clot outside surface and the MERD interior surface are set as the slave and master surfaces, respectively; and (2) the arterial inner wall is given priority as the master surface, while the embolus's outer surface is defined as the slave surface. The symmetrical section of the embolus is also restrained circumferentially. Also, an axial displacement is applied to the proximal tip of the device to achieve system migration. The other FEA module definitions of interactions and boundary conditions remain as previously described.
To validate the mesh method and algorithmic analysis, a circular symmetrical diamond-shaped pattern is generated for the numerical simulation and experimental test. In the radial force test (as shown in Fig. 4), a desktop system (Electromechanical test system of RX650 (MSI, Phoenix, AZ)) is applied to compress the tubular device from OD1 (outer diameter) to OD2 at the crimpling speed 0.1 mm/s under 37 °C while force value is measured. The real-time force is obtained gradually from the test system sensor. The radial force (RF) value (the vertical axis) is plotted versus the outer diameter of the MERD (the horizontal X axis) in Fig. 5.
In the process of numerical simulation, the crimping tool is replaced by a rigid shell to compress the original pattern. The contact properties, the modeling method, and the boundary condition are designed similarly to the shape-setting, crimping, deployment, and migration statuses. The targeted radial force is measured and obtained from the reactive force applied on the crimping cylinder.
Figure 5 indicates that the test and the simulation curves are similar to each other; that is, all of the numerical outcomes are relatively accurate and correct, which should provide confidence in the prediction of the performance analysis and structural optimization.
Expansion and Annealing.
Figure 6 shows the radial profile expansion and axial length shrinkage of the MERD at the top of the figure. With regard to the macrostructure, the MERD's outer diameter increases from 2.00 to 4.00 mm, while the axial length decreases from 26.9 to 21.7 mm with a shrinkage ratio of 21.7%. In contrast to the conventional metal materials, the evaluated mechanical behavior of Nitinol is strain-induced. Cracking or fracturing may occur in the “failure-prone” region due to an excessive MERD expansion. Figure 6 also shows the MPS distribution of the expanded device.
It is observed that high-magnitude strain is located in the fillet regions of the struts' internal side (the maximum expanding strain is 9.0%). The strain value in the rest of the regions is found to be negligible. Compared to the Nitinol strain threshold value of 12.0% , this peak value is considered acceptable. The length shrinkage and strain concentration are assumed to occur due to axial compression of the tubular lattice, which is induced by the expansion effect. Finally, the annealing step is used to achieve a stress/strain-free state and accomplish the final set shape.
The mechanical crimping step is predicted to be the most critical procedure for the Nitinol embolus retrieval device. High-magnitude crimping strain may result in undesired fractures or cracks in the structure. The semiclosed tubular pattern consists of a distal colander-shaped “tail prong” and the principal mesh. The comparison of the structure configuration between uncrimped and crimped MERDs is shown in Figs. 7(a) and 7(b). During the crimping, there is an axial elongation of the tubular mesh structure. Figure 7(c) shows the MPS contour plots of the strain pattern when the MERD is shrunk into the catheter of 0.5 mm OD. It is observed that high-magnitude strains are distributed more extensively than in the shape-setting step.
As shown in Fig. 7(c), high-magnitude strains are concentrated in the “shoulder” of the strut's external side. The strain peak value of 12.1% is located in the flaw-prone region of the MERD's proximal slope head. The maximum strain exceeds the strain limit of Nitinol material; that is, a crack or fracture may occur during the progressive crimping procedure. The elongation and the strain field are found to occur due to the axial tension of the tubular pattern as the crimping-induced tensile force is generated to stretch the mesh cells axially.
The deployment simulation outcome is presented to investigate the effectiveness and safety of the embolus retrieval device system. Figure 8(a) shows the Von Mises stress (VMS) contour plots of the expanded blood vessel wall.
Due to the support of tubular struts, the artery deforms significantly and encounters a stress concentration. As a result, the peak stress is located in the regions which make contact with the connecting struts between the principal part and the colander-shaped tail. In 0.624 s of the deployment procedure, the peak stress of the blood vessel wall reaches 0.309 MPa. Based on a yield stress of 0.67 MPa reported for abdominal aortic aneurysms (which may also be pertinent here) , the expanded artery is suggested to be secure; that is, it has minimal risk of serious trauma or rupture clinically. For a peak stress reduction, geometrical and parametric optimization for the connecting struts can be proposed in the future. The strain field contour plots of the tubular MERD are shown in Fig. 8(b). After the MERD self-expands and penetrates into the blood vessel wall, the strut's exterior sides (or so-called shoulder) exhibit high-magnitude strains. The peak MPS value of 6.3% is considered acceptable in engineering terms. It is observed from Fig. 8(c) that the tubular mesh device can be deployed and inserted into the nominal artery completely and stably.
A schematic of the strain distribution of the MERD during the MERD system migration is shown in Fig. 9(a).
The FEA analysis is expected to help understand the influence of embolus retrieval along cerebral vessels. In this step, a bulletlike plaque embolus is added into the tubular mesh structure. It is observed from Fig. 9(b) that the MERD can be withdrawn so that it migrates along the cerebral vessel smoothly and stably. The peak stress of the artery wall in the embolus retrieval procedure is 0.328 MPa. In contrast to the previous research, the artery has little risk of injury or rupture during operation. The peak MPS of the compressed MERD is 6.5%. As an operation instrument, the strain field does not exceed the limits of Nitinol material and is predicted to be safe during embolus retrieval. It is found that the embolus clot can be captured by the semiclosed mesh structure successfully. The embolus-dominated contact effect for the Nitinol pattern shown is negligible. The stress field distribution plot of the plaque embolus is shown in Fig. 9(c). The peak strain is located on the exterior surface where contact with the colander-shaped struts occurs. Finally, it is considered that the semiclosed multilattice tubular MERD has good evidence capture plaque emboli successfully and smoothly, without clinical accidents or device structure failure.
The results demonstrate the mechanical performance of a MERD during physiological loading conditions including shape-setting, crimping, deployment, and embolus capture. FEA is used to simulate and analyze all steps of the physical stent deployment process, thereby showing that maximum stress is not exceeded at any stage.
The friction factor, a stabilization factor, nonlinear analysis, and penalty contact property settings are used to improve convergence. For the shape-setting procedure, penalty contact property and nonlinear analysis settings are used to solve the contact interaction equations. To maintain equilibrium, the static step module of abaqus/standard is chosen. A friction factor of 0.2 is used to reflect sliding contact effects. A stabilization factor is also applied to aid convergence due to loss of contact. At the end of the shape-setting, annealing is simulated to achieve a stress-free state and accomplish the final shape-setting configuration. The penalty contact method is selected for the contact property definitions. For the crimping procedure, a friction coefficient of 0.25  is used to obtain contact effects. A stabilization factor of 0.1 is selected to improve convergence and dampen contact vibration. A smooth or gradual contact relationship is established between the embolus retrieval device and the crimping tool. At the end of this step, the MERD is completely compressed into a rigid catheter.
Based on the analysis above, a number of recommendation are suggested for engineers for optimization of MERD structures: In the step of shape-setting, the high-magnitude strain reduction of a MERD can be achieved by double expansion and annealing. For the crimping procedure, the geometry of the flaw-prone region should be optimized. To reduce the peak strain, the strut's external side (the so-called shoulder region) and internal side (close to the fillet curve region) should be smooth and continuous. Also, the radius of curvature of the fillet region should also be adjusted. To reduce the peak arterial stress during insertion of the embolus-retrieval device, the MERD radial force should be reduced. Thus, a decrease in strut thickness and width is proposed. In order to decrease the stress and axial length fluctuation, the topology geometry of the connecting bridge could be optimized. The particular structure not only meets technical requirements of MERD but also provides a reference for a future similar geometric design. As described above, the pitched proximal end would contain the thrombus without disturbance. In addition, the “mouth” sharplike segment may be easier to be dragged into a low-profile catheter. The diamond-shaped meshes will provide relative low support to the artery, allowing thrombus to run into semiclosed area. The distal end, shaped like a cage, can be used to capture debris, avoiding particles escaping downstream.
In future work, the cerebral artery will be modeled on the basis of magnetic resonance imaging images for an accurate 3D reconstruction instead of an abdominal aortic aneurysms artery material property. A comparison of the simulated effects of differing topology and dimensions of MERDs will be presented. The rigid crimping tool will be modified by using a deformable nylon catheter. The delivery procedure of the embolus retrieval system will also be demonstrated.
During interventional mechanical thrombectomy, the management of the delivered system deployment and embolus capture are considered crucial and difficult to optimize. The method provided in the paper is a suggested as an operational guideline for modeling accurate MERD insertion during clinical use. Based on the demonstrated axial shrinkage during MERD self-expansion, physicians can deliver more confidently the embolus device system to the correct position in a diseased cerebral vessel. In this way, the MERD will self-expand and capture plaque emboli more reliably. The simulation results can be utilized to verify the performance of the delivery system insertion and the effectiveness of embolus retrieval.
The third author was partially supported by the National Science Foundation (IIS-1251069) and the National Institutes of Health (1R01EB018205-01).