Volume 21, Issue 3 , Pages 367-374, March 2010
Toward an Optimal Position for Inferior Vena Cava Filters: Computational Modeling of the Impact of Renal Vein Inflow with Celect and TrapEase Filters
Article Outline
Purpose
To evaluate the hemodynamic effects of renal vein inflow and filter position on unoccluded and partially occluded inferior vena cava (IVC) filters with use of three-dimensional computational fluid dynamics.
Materials and Methods
Three-dimensional models of the TrapEase and Günther Celect IVC filters, spherical thrombi, and an IVC with renal veins were constructed. Hemodynamics of steady-state flow was examined for unoccluded and partially occluded TrapEase and Günther Celect IVC filters in varying proximity to the renal veins.
Results
Flow past the unoccluded filters demonstrated minimal disruption. Natural regions of stagnant/recirculating flow in the IVC were observed superior to the bilateral renal vein inflows. High flow velocities and elevated shear stresses were observed in the vicinity of renal inflow. Spherical thrombi induce stagnant/recirculating flow downstream of the thrombus. Placement of the TrapEase filter in the suprarenal position resulted in a large area of low shear stress/stagnant flow within the filter just downstream of thrombus trapped in the upstream trapping position.
Conclusions
Filter position with respect to renal vein inflow influences filter trapping hemodynamics. Placement of the TrapEase filter in a suprarenal location may be thrombogenic, with redundant areas of stagnant/recirculating flow and low shear stress along the caval wall caused by the upstream trapping position and the naturally occurring region of stagnant flow from the renal veins. Infrarenal vein placement of IVC filters in a near-juxtarenal position with the downstream cone near the renal vein inflow likely confers increased levels of mechanical lysis of trapped thrombi from increased shear stress from renal vein inflow.
Abbreviations: IVC, inferior vena cava, Re, Reynolds number, WSS, wall shear stress
INFERIOR vena cava (IVC) filters have played an integral role in the prevention of pulmonary embolism from deep vein thrombosis for more than 30 years. More than 100,000 filters are placed annually in the United States alone (1). Despite this extensive use of filters, very little is understood about the ideal position for an IVC filter. Currently, IVC filter manufacturers have received approval for filter placement only in the infrarenal IVC, though off-label use in the suprarenal IVC with particular clinical scenarios (eg, thrombus in the gonadal or renal veins or within the IVC at or near the level of the renal veins, anatomic variation in the IVC and renal vein, and intrinsic or extrinsic IVC narrowing) is well documented (2).
An ideal IVC filter traps significant thrombus without significantly increasing the risk of thrombosis from inducing a prothrombotic state or trapping emboli. Indeed, the PREPIC (Prévention du Risque d'Embolie Pulmonaire par Interruption Cave) study group (3) noted a statistically significant increase in deep vein thrombosis 8 years after placement of a permanent IVC filter and suggested that this may be related to thrombosis at the filter site. A causal relationship between deep vein thrombosis and filters has not been established, but two potential theories are that (i) filters may cause progressive damage to the vein wall with secondary IVC occlusion at the filter or narrowing and eventual stenosis that contributes to deep vein thrombosis at or below the filter or (ii) IVC filters may induce a local prothrombotic state either by their design or after trapping emboli (4). Most endovascular interventionalists advocate filter placement in an infrarenal position, thereby decreasing the risk of renal vein thrombosis from potential filter occlusion. The “infrarenal” portion of the IVC is a vague term that includes the entire length of the IVC from the renal veins caudally to the IVC bifurcation into the common iliac veins. With respect to renal vein inflow, where is the precise, ideal position to deploy an IVC filter? Does such a location exist?
The Society of Interventional Radiology Foundation consensus panel on the development of a research agenda for IVC filters recently published a list of basic science research priorities, which includes “computerized flow dynamic studies of multiple filters” (5). Both in vitro bench experiments and computational flow modeling of several types of filters have been reported (4, 6, 7, 8). However, we are aware of no study that has incorporated renal vein inflow. Building on the work of Singer et al (4), which evaluated the TrapEase IVC filter (Cordis, Miami Lakes, Florida) in a computational flow model and corroborated results with bench experiments, the present study models the TrapEase and Günther Celect filters (Cook, Bloomington, Indiana) with and without nonocclusive thrombus in an IVC model that incorporates anatomically correct renal vein inflows. Computer simulations were run with the filters in varying proximity to the renal veins to determine whether an ideal position exists with regard to renal vein inflow and filter flow dynamics.
Materials and Methods
Three-dimensional computer models of the IVC with renal veins, Günther Celect filter, TrapEase filter, and simulated thrombi were constructed to study flow dynamics with the filters in various positions. Hemodynamic properties in and around unoccluded and partially occluded filters were examined via three-dimensional computational fluid dynamics.
IVC and Renal Vein Model
As previously described by Singer et al (4), the IVC was modeled as a rigid, straight pipe with a diameter of 23 mm, per the average IVC diameter described by Kaufman et al (9). Renal veins were also modeled as straight rigid pipes.
The diameters and entry angles of the renal veins into the IVC were calculated from abdominal computed tomography (CT) scans of 24 patients (12 male, 12 female; average age, 55 years, age range, 16–89 y) with no known renal disease or variant renal venous anatomy. Renal vein diameters were measured from intravenous contrast medium–enhanced axial CT images (2.5-mm slice thickness; Lightspeed 16; GE Medical Systems, Milwaukee, Wisconsin) near the confluence with the IVC. Renal vein entry angles were measured using picture archiving and communication system software (Stentor I-site, version 3.3.1; Philips, Best, The Netherlands) from intravenous contrast medium–enhanced coronal CT images (5-mm slice thickness) reformatted from the axial source images. Coronal reformatted images were also used to identify which renal vein was more caudal in location, and the craniocaudal distance between the right and left renal veins was recorded. Average bilateral renal vein diameters (right, 10.5 mm ± 2.0; left, 8.8 mm ± 1.8), bilateral renal vein entry angles (right, 62.0° ± 26.8°; left, 60.7° ± 11.1°), and cranial–caudal distance between the renal vein entry levels (6.8 mm ± 7.5) were used to construct the model of the IVC and renal veins (Fig 1). The right renal vein was more caudal than the left in a majority of the patients (right more caudal, n = 10; left more caudal, n = 3; approximately the same level, n = 11). Therefore, the model was constructed with the right renal vein entry 10 mm more caudal than the left.
Figure 1.
Schematic diagram of the three-dimensional model showing the IVC, renal veins, and filter positions. Letters A–F correspond to positions of the downstream cone of the Celect and TrapEase filters. Positions A and B are infrarenal, C is just below juxtarenal position with the cone near the level of the renal vein inflow, D and E are juxtarenal, and F is suprarenal in position.
Patient anatomic information used from the CT scans was given exemption status after meeting our institutional review board and federal regulatory criteria for exemption.
TrapEase and Günther Celect Filter Models
Construction of the TrapEase model was described by Singer et al (4). The computer model of the Günther Celect filter was similarly constructed according to methods of computer-aided design. In particular, the Celect filter was inserted into a glass test tube with an inner diameter of 23 mm, and high-resolution digital photographs were taken with a DiMAGE-Xt digital camera (Minolta, Osaka, Japan). Measurements of the filter were also obtained with a Cen-tech 6-inch digital caliper (Harbor Freight Tools, Camarillo, California). The photographs and measurement data were then imported into the GNU Image Manipulation Program (GNOME Foundation, Groton, Massachusetts), in which spatial geometry of the filter was extracted on the basis of pixel color and location. The geometric specifications were imported into the Overture software framework (Lawrence Livermore National Laboratory, Livermore, California) (10, 11), in which computer models were constructed for each filter. For ease in modeling, the extraluminal barbs were excluded from the model because their presence does not alter the characteristics of flow intraluminally.
Clot Models
As previously described by Singer et al (4), spherical thrombi were modeled as rigid spheres. The volumes were 0.5 mL and 1.0 mL, which is consistent with the visual scale described by Wang et al (12) and similar to the volumes used in previous studies of the TrapEase filter.
Simulations
As previously described by Singer et al (4), flow was modeled as an incompressible, Newtonian fluid whose motion is described mathematically by the Navier–Stokes equation (13). The Navier–Stokes equations were solved with the incompressible flow solver within the Overture software framework, and postprocessing was performed using tools provided by Overture, custom scripts written in Matlab (MathWorks, Natick, Massachusetts), and the GNU Image Manipulation Program.
Simulations were performed with unoccluded and partially occluded TrapEase and Günther Celect filters. Thrombi were positioned in the upstream and downstream trapping positions of the TrapEase filter and in the single downstream position of the Günther Celect filter. Both filters were placed in the geometric center of the simulated cylindrical IVC and were placed in varying locations in the cranial–caudal plane of the IVC to simulate infrarenal vein placement (Fig 1, positions A and B), juxtarenal vein placement (Fig 1, positions C–E), and suprarenal vein placement (Fig 1, position F).
The mean inlet velocity of the infrarenal vein IVC was 3.44 cm/sec, which corresponds to a flow rate of 0.86 L/min in the 23-mm-diameter IVC and a Reynolds number (Re) of 320 (Re = ρUD / μ, where ρ is the density of blood [1,040 kg/m3], U is the mean inlet velocity, D is the diameter of the vena cava [23 mm], and μ is the viscosity of blood [2.57e-3 kg/msec]). Flow at this Re is laminar. Even though a Re of 600 has been used in previous studies (4, 6, 8, 14), the corresponding flow rate (e.g., 2 L/min in a 2-cm vena cava [6]) is more indicative of higher flow velocities typically seen in the suprarenal IVC. The peak renal vein flow velocities were 40.2 cm/sec (0.73 L/min, Re = 715) and 43.5 cm/sec (1.13 L/min, Re = 925) for the right and left veins, respectively. These velocities are consistent with previous reports of Doppler renal vein velocities in normal adults (15). Inflow to the IVC and renal veins was specified by parabolic velocity profiles.
Results
As in Singer et al (4), the three-dimensional Navier–Stokes equations were solved for the velocity and the pressure. In the contour plots of the axial velocity, all velocities were normalized by the corresponding value for fully developed pipe flow. Because the speed of renal inflow is significantly greater than IVC flow (as noted earlier), the color scales show the greatest variation of flow speeds near the renal veins: flow upstream of the renal veins is relatively uniform. The wall shear stresses (WSSs) were normalized by the corresponding value for pipe flow, and the breaks in the line plots denote the locations of renal inflow, where no wall is present. In all figures showing flow and normalized velocities, IVC flow is from the bottom to the top, and renal vein inflow is toward the vena cava.
Unoccluded Filters
For the unoccluded TrapEase and Celect filters (Fig 2), there is minimal disruption to the flow upstream of the renal veins. That is, Figure 3, Figure 4 show that the normalized WSS upstream of the renal veins is nearly constant and equal to unity, which corresponds to pipe flow. In addition, on the velocity scales that capture the dynamics of the renal inflow, both filters demonstrate only small deviation from pipe flow, and renal inflow impacts only the flow field downstream of the renal veins. Consequently, the flow upstream of the renal veins is comparable to that studied by Singer et al (4), in which the unoccluded TrapEase filter did not disrupt the flow significantly. Near the sites of renal inflow, the renal veins act as jets that introduce high-speed flow into the vena cava, and the flow downstream of the renal veins is disrupted. Immediately downstream of both renal veins, regions of low velocity and recirculating flow are observed near the wall of the cava, which is a result of the large velocity difference between IVC flow and renal flow (arrows, Fig 2). The filters disrupt the renal inflow in close proximity to the renal veins; the flow must change direction (thereby increasing the transverse components of velocity) to bypass the filter and flow downstream. When the filter is proximal to or downstream of the renal veins, flow inside the filter is disrupted significantly (Fig 2).
Figure 2.
Models of the unoccluded TrapEase and Celect filters within the computational model of the IVC and renal veins. Filters are positioned at vena cava positions C and F in the schematic diagram in Figure 1. Note the regions of naturally occurring stagnant flow (arrows) along the bilateral vena cava walls just downstream of the renal venous inflow.
Figure 3.
Normalized WSS profiles for the Celect filter in different IVC positions (see A–F in Fig 1) with a 1.0 mL spherical thrombus in the central downstream trapping position. Dotted line indicates Celect filter at position C without thrombus. Both sides of the vena cava wall are evaluated. Flow is from left to right, and the x-axis denotes the distance (in mm) from the downstream tip of the filter. Symmetry is observed in the Celect filter with low WSSs downstream of the thrombus and relatively high shear stresses with suprarenal placement (position F in Fig 1). Breaks in the WSS lines correspond with the locations of renal vein inflow where no wall is present.
Figure 4.
Normalized WSS profiles for the TrapEase filter in different IVC positions (see A–F in Fig 1) with a 1.0 mL spherical thrombus in the central downstream trapping position (a) or a 0.5 mL spherical thrombus in the right lateral upstream trapping position (b). Dotted line indicates TrapEase filter at position C (see Fig 1) without thrombus. Both sides of the vena cava wall are evaluated. Flow is from left to right, and the x-axis denotes distance (in mm) from the downstream tip of the filter. Breaks in the lines of WSS correspond with the locations of renal vein inflow where no wall is present. (a) Symmetry is observed in the TrapEase filter with low WSSs downstream of the thrombus and relatively high shear stresses with suprarenal placement (position F in Fig 1). (b) At position F (red line), there is a decrease in the WSS along the right wall at approximately 50 mm proximal to the downstream tip of the filter, which is a result of the overlap between the stagnant/recirculating flow downstream of the renal vein inflow (from 55 mm to 60 mm upstream of the tip) and the flow disruption caused by the upstream, laterally trapped thrombus.
The WSS profiles for both unoccluded filters exhibit regions of low WSS immediately downstream of the renal inflow (dotted lines in Figure 3, Figure 4). As noted earlier, these areas correspond to regions of low-velocity flow that is pushed aside by high-velocity flow coming from the renal veins. Consequently, the recirculating flow gives rise to negative velocity gradients, which produce large, negative WSSs. Upstream of the renal inflow, the WSSs are nearly uniform, and the normalized values are close to unity, thereby indicating minor deviations from pipe flow.
Partially Occluded Filters
Partial occlusion of the TrapEase and Celect filters disrupts flow downstream of the thrombus (Figure 5, Figure 6). However, when trapped thrombus is proximal to the renal veins, incoming renal flow is redirected by the thrombus and forced downstream along the vena cava wall. Consequently, flow along the vena cava wall (immediately downstream of renal inflow) is accelerated, and the volume of stagnant/recirculating flow is reduced compared with the unoccluded configuration. In addition, the renal vein inflow may be directed at the thrombus trapped downstream when the filter cones are positioned closer to the renal veins.
Figure 5.
TrapEase filter at IVC positions A, C, and F (see Fig 1) with a 0.5 mL thrombus in the upstream trapping position (A1, C1, and F1) and a 1.0 mL thrombus in the downstream trapping position (A2, C2, and F2). Color scale corresponds with normalized velocities. For configurations F1 and F2, the arrows denote regions of naturally occurring stagnant/recirculating flow caused by renal inflow.
Figure 6.
Flow past the Celect filter partially occluded with a 1.0-cm3 thrombus. The velocity color scale is normalized (to pipe flow), and all images have the same scale. The filters are positioned at IVC stations A–F (see Fig 1).
As previously described by Singer et al (4), partially occlusive thrombus in the upstream trapping position of the TrapEase filter results in a region of low-speed, intrafilter flow (Fig 5) that is accompanied by low shear stresses along the ipsilateral vena cava wall (Fig 4). This region of stagnant/recirculating flow can be larger when the TrapEase filter is placed in a suprarenal location, as the stagnant region along the vena cava wall caused by the nonocclusive thrombus overlaps with a naturally occurring region of stagnant/recirculating flow just downstream to the renal vein inflow bilaterally (Fig 5).
WSS profiles for the two filters are qualitatively similar (Figure 3, Figure 4). In particular, for the filters upstream or inferior to the renal veins, the peak WSS occurs near the narrow passage between the vena cava wall and the trapped thrombus; the minimum WSS is near the stagnant/recirculating zone immediately downstream of renal inflow. When the filters are in a suprarenal vein location, the locations of maximum and minimum wall stress remain unchanged (near the thrombus and downstream of renal inflow, respectively), but the trapped thrombus is downstream of renal inflow, which leads to a comparatively large peak in the WSS as a result of the high-speed flow past the thrombus.
Discussion
The present computational flow study indicates that renal vein inflow has significant hemodynamic effects on blood flow near IVC filters. Based on previous magnetic resonance (MR) imaging and ultrasound (US) studies that included groups of control patients (15, 16, 17), there is some variability in peak renal venous velocities. The renal inflow velocities used here fall well within the parameters from earlier physiologic studies. The renal venous flow in the model dominates the overall flow pattern in the IVC downstream from the renal inflow. This dominant flow is perhaps best demonstrated by a naturally occurring region of stagnant/recirculating flow, which is superior to the confluence of each renal vein (arrows, Figure 2, Figure 5).
Singer et al (4) previously demonstrated intrafilter regions of low shear stress and stagnant/recirculating flow along the vena cava wall, immediately downstream of the upstream trapping position of the TrapEase filter. Regions of low shear stress and stagnant/recirculating flow can be thrombogenic as a result of the accumulation of thrombin and fibrin (18). Recent clinical studies suggest a higher incidence of intrafilter thrombus and caval thrombosis with the nearly structurally identical OptEase filter as described by Singer et al (4). The upstream trapping position is particularly concerning with the TrapEase filter in a suprarenal position, as the region of stagnant/recirculating flow just downstream of trapped thrombus in the upstream trapping position can overlap with the naturally occurring region of stagnant/recirculating flow just superior to the renal vein confluence bilaterally (Fig 5). If the two regions overlap, the portion of vessel occupied by stagnant flow is larger and may increase the risk for thrombosis along the caval wall.
The present study presents results from a computational flow model of the Günther Celect filter with renal inflow. In its optimal position, which is centered in the IVC, the Celect filter demonstrates minimal flow disturbance in the unoccluded state. Since the Celect filter has only one trapping position, the resulting flow dynamics are similar to the downstream trapping position of the TrapEase filter. Both partially occluded filters demonstrated similar regions of low shear stress and stagnant flow downstream from the simulated thrombi (Figure 5, Figure 6). High shear stress is seen along the caval wall downstream from the thrombus in the infrarenal filter positions. As both filters are moved more superiorly, with the downstream capture cone just below or near the level of the renal vein inflow (Fig 1, position C), the volume of stagnant/recirculating flow downstream from the filter decreases. In addition, high velocities and shear stresses are seen along the downstream filter cone. These conditions may reduce the risk of primary hemostasis by stimulating the secretion of tissue plasminogen activator and therefore reduce the risk of secondary hemostasis by clearing fibrin and thrombin and increasing mechanical lysis of thrombi (19). Therefore, placement of these two IVC filters in a near juxtarenal location (Fig 1, position C) may decrease the risk of thrombus propagation within the filter cone because significantly higher shear stresses and velocities from renal vein inflow serve to reduce stagnant flow and may improve mechanical lysis of trapped thrombi.
Positioning the filters in a true juxtarenal position (Fig 1, position D) results in similar regions of high shear stresses and velocities in the central, downstream trapping positions of both filters. Even though these flow conditions may also serve to mechanically lyse smaller thrombi trapped in the filter, positioning the filter at this level must be weighed against the risk of caval occlusion (possibly from large, unstable lower extremity deep vein thrombus) and potential subsequent bilateral renal vein occlusion. In addition, the legs and centering struts of the Celect filter may become engaged in the renal veins.
Suprarenal placement of the filters demonstrates the dominant flow of the renal veins relative to flow through the IVC. High shear stresses and velocities are seen along the caval walls in the partially occluded TrapEase and Celect filters in the downstream positions. In the case of partially occluded filters, the high flow rates and shear stresses along the caval walls suggest a decreased risk of secondary hemostasis as a result of mechanical lysis and clearing of fibrin and thrombin. A recent retrospective, single-center review of suprarenal filters (N = 70) by Kalva et al (2) described no patients who presented with caval thrombosis during a mean follow up of 573 days; however, it is worth noting that, in the 30 patients who had follow-up abdominal CT scans, intrafilter thrombus was noted in three cases (10%), two with a TrapEase filter and one with a Günther Tulip filter (2). In addition, the thrombus was “seen in the periphery of the TrapEase filter as a thin rim attached to the wall of the IVC. In the case of the Tulip filter, the thrombus was seen in the apex of the filter” (2). These clinical observations regarding the location of trapped thrombus correlate well with results from our flow models. Specifically, trapped thrombus in the upstream trapping position of the TrapEase filter may induce a larger region of stagnant flow along the ipsilateral vena cava wall as a result of a naturally occurring region of stagnant flow that is already present because of renal inflow. This could lead to intrafilter thrombus along the vena cava wall, which may be difficult to clear because of stagnant flow. This correlates with what was described clinically by Kalva et al (2).
Our computer model is founded on simplified, but generally realistic, assumptions. The walls of the vena cava and renal veins are modeled as smooth and rigid even though in vivo the vena cava and renal veins are dynamic and demonstrate variations in angle and radial diameter. Our model is also limited in that the vena cava typically has a larger diameter in the suprarenal portion, but for simplicity the infrarenal and suprarenal IVC are taken to have a uniform diameter. Our steady-state flow velocities from the renal veins and IVC are based on clinical studies from MR imaging and US (15, 16, 17), but in vivo the inflow velocities are unsteady and change with exercise and physiologic conditions. This study is limited to spherical thrombi, but based on results from Singer et al (4), different shapes generally result in similar shear stresses and velocity profiles, especially when thrombi are trapped in the center of the IVC. Swaminathan et al (20) also noted that spherical thrombi represent, in some sense, a statistical average of irregular shapes. The diameters and angles of the renal veins are based on average values from patient CT scans, even though significant variation is seen within the general population and renal venous anatomic variants are common (eg, circumaortic left renal vein, retroaortic left renal vein, and duplicated IVCs) (21). Other sources of IVC inflow including lumbar veins and other venous tributaries are not included in our model. Finally, the Celect filter was modeled only in the center of the IVC, but the Celect is susceptible to tilting at the time of deployment (22).
Our computational flow model of the IVC and renal veins suggests that the ideal location for infrarenal IVC filter placement is immediately upstream of the juxtarenal position, with the downstream cone of the filter near the level of renal inflow (Fig 1, position C). The dominating high-velocity inflow from the renal veins serves as a source of higher shear stresses and flow velocities that may decrease primary and secondary hemostasis, particularly in the case of nonocclusive thrombi trapped in the downstream filter cone. True juxtarenal positioning confers similar higher shear stresses but also increases the risk of renal vein thrombosis from filter occlusion. Suprarenal placement of IVC filters may confer a decreased risk of hemostasis as a result of much higher WSS and flow velocities; however, in the event of a large occlusive thrombus, the risk of renal vein thrombosis must be considered. Suprarenal placement of the TrapEase filter carries the additional risk of inducing large, redundant areas of stagnant flow at the upstream trapping position that may be thrombogenic. Given the results of this computational flow model, future clinical studies of IVC filters that examine outcomes based on the relative position of filters with regard to renal veins are warranted.
References
- Lee P, Raizada A, Ciocca R. Growing utilization of IVC filter placement from 2001-2005: analysis of NIS. Presented at the Society for Clinical Vascular Surgery 2009 Meeting; March 18–21, 2009; Fort Lauderdale, FL.
- Suprarenal inferior vena cava filters: a 10-year experience. J Vasc Interv Radiol. 2008;19:1041–1047
- . Eight year follow-up of patients with permanent vena cava filters in the prevention of pulmonary embolism. Circulation. 2005;112:416–422
- . Computational modeling of blood flow in the TrapEase inferior vena cava filter. J Vasc Interv Radiol. 2009;20:799–805
- Development of a research agenda for inferior vena cava filters: proceedings from a multidisciplinary research consensus panel. J Vasc Interv Radiol. 2009;20:697–707
- . Hemodynamic effects of clot entrapment in the TrapEase inferior vena cava filter. J Vasc Interv Radiol. 2004;15:485–490
- Effects of thrombosed vena cava filters on blood flow: flow visualization and numerical modeling. Ann Biomed Eng. 2008;36:1764–1781
- . Vena cava filter performance based on hemodynamics and reported thrombosis and pulmonary embolism patterns. J Vasc Interv Radiol. 2007;18:103–115
- . Anatomical observations on the renal veins and inferior vena cava at magnetic resonance angiography. Cardiovasc Interv Radiol. 1995;18:153–157
- . An object oriented framework for solving partial differential equations. In: Scientific computing in object-oriented parallel environments: Springer lecture notes in computer science. vol. 1343:New York: Springer; 1997;p. 177–194
- . Moving overlapping grids with adaptive mesh refinement for high speed reactive and non-reactive flow. J Comp Phys. 2006;216:744–779
- . Estimation of trapped thrombus volumes in retrievable inferior vena cava filters: a visual scale. J Vasc Interv Radiol. 2007;18:273–276
- . Incompressible flow. In: Incompressible flow. 2nd ed.. New York: John Wiley and Sons; 1996;p. 228–253
- . In vitro hemodynamic evaluation of a Simon nitinol vena cava filter: possible explanation of IVC occlusion. J Vasc Interv Radiol. 2001;12:613–618
- Noninvasive diagnosis of renal thrombosis by ultrasonic echo-Doppler flowmetry. Kidney Int. 1983;23:882–887
- . Inferior vena cava hemodynamics quantified in vivo at rest and during cyclic exercises using magnetic resonance imaging. Am J Physiol Heart Circ Physiol. 2003;284:1161–1167
- Hemodynamic investigation of the left renal vein in pediatric varicocele: Doppler US, venography, and pressure measurements. Radiology. 2006;241:228–234
- . Virchow's triad revisited: abnormal flow. Pathophysiol Haemost Thromb. 2004;33:455–457
- . Platelets and shear stress. Blood. 1996;88:1525–1541
- . Numerical analysis of the hemodynamics and embolus capture of a Greenfield vena cava filter. J Biomed Eng. 2006;128:360–370
- Congenital anomalies of the inferior vena cava and left renal vein: evaluation with spiral CT. J Vasc Interv Radiol. 1998;9:339–345
- Deployment performance and retrievability of the Cook Celect vena cava filter. J Vasc Interv Radiol. 2009;20:376–383
Computer time on the Lawrence Livermore National Laboratory's Yana cluster was provided under Livermore Computing's Multiprogrammatic and Institutional Computing Initiative. The Lawrence Livermore National Laboratory is operated by Lawrence Livermore National Security for the United States Department of Energy, National Nuclear Security Administration, under Contract DEAC5207NA27344.
Neither author has identified a conflict of interest.
PII: S1051-0443(09)01145-2
doi:10.1016/j.jvir.2009.11.013
© 2010 SIR. Published by Elsevier Inc. All rights reserved.
Volume 21, Issue 3 , Pages 367-374, March 2010