In vitro angiographic comparison of the flow-diversion performance of five neurovascular stents

Abstract

Background and purpose

Data differentiating flow diversion properties of commercially available low- and high-porosity stents are limited. This in vitro study applies angiographic analysis of intra-aneurysmal flow to compare the flow-diversion performance of five neurovascular devices in idealized sidewall and bifurcation aneurysm models.

Methods

Five commercial devices (Enterprise, Neuroform, LVIS, FRED, and Pipeline) were implanted in silicone sidewall and bifurcation aneurysm models under physiological average flow of blood analog fluid. High-speed angiographic images were acquired pre- and post-device implantation and contrast concentration-time curves within the aneurysm were recorded. The curves were quantified with five parameters to assess changes in contrast transport, and thus aneurysm hemodynamics, due to each device.

Results

Inter-device flow-diversion performance was more easily distinguished in the sidewall model than the bifurcation model. There were no obvious overall statistical trends in the bifurcation parameters but the Pipeline performed marginally better than the other devices. In the sidewall geometry, overall evidence suggests that the LVIS performed better than the Neuroform and Enterprise. The Pipeline and FRED devices were statistically superior to the three stents and Pipeline was superior to FRED in all sidewall parameters evaluated.

Conclusions

Based on this specific set of experiments, lower-porosity flow diverters perform significantly better in reducing intra-aneurysmal flow activity than higher-porosity stents in sidewall-type geometries. The LVIS device is potentially a better flow diverter than the Neuroform and Enterprise devices, while the Pipeline is potentially better than the FRED.

Introduction

In the last decade, flow diversion has emerged as an endovascular treatment option for complicated aneurysms such as wide-necked, giant, and fusiform aneurysms that are not amenable to coiling.^1–8 Flow-diverting devices are deployed across the neck of the aneurysm to impede flow inside the aneurysm and provide a scaffold for cellular proliferation and development of a layer of endothelium over it. The process of parent vessel remodeling^3,8,9 and establishment of a neointima excludes the aneurysm from the circulation. Numerous experimental, computational fluid dynamics (CFD), and angiographic studies have been conducted to evaluate the reduction in intraneurysmal flow activity due to stent deployment in the parent vessel. ¹⁰ The efficacy of flow diverters in reducing intraneurysmal flow depends on porosity and pore density, or permeability, of the device as well as the anatomic configuration of the aneurysm and its positioning relative to the oncoming flow (geometry of the aneurysm-parent vessel complex).^3,9–18 In general, it is well established that lowering the device porosity and increasing the device pore density results in greater reduction of intraneurysmal flow activity.^{11,13,16,19–26} The primary clinical utility of all these studies is to be able to predict device efficacy (long-term aneurysm occlusion) based on intraneurysmal flow reduction quantities. Sample sizes are sparse, but several studies based on clinical data^27–33 collectively note that the reduction in intraneurysmal flow after flow diversion can be significantly different between aneurysms that remain patent at follow-up versus those that occlude at follow-up; the sensitivity and specificity of these parameters seem reasonably high at 75%–90%.^28,29,31,32 Treatment with flow diverters is used for complicated aneurysm geometries not amenable to coiling. In certain instances, when a flow diverter is not approved or indicated for treatment of a wide-necked or fusiform aneurysm, high-porosity stents may be deployed for stent-assisted coiling. The stents themselves are postulated to provide some flow-diversion assistance along with a scaffold for aneurysm neck remodeling.

As angiography is utilized for all endovascular treatment, the adequacy of flow diversion following deployment of the first device can be based on qualitative angiographic assessment of contrast transport into and out of the aneurysm. Increase in contrast residence time in the aneurysm may be assessed based on contrast recirculation or pooling within the aneurysm or a slower washout of contrast from the aneurysm compared to the contrast washout time prior to device implantation. If needed, more devices can be deployed until contrast residence time in the aneurysm is deemed to be sufficient to occlude the aneurysm, which is confirmed by follow-up examination. In case the aneurysm is not completely occluded at follow-up, further treatment options can be decided at that time.^1,4

Angiography has been around for nearly 90 years and several studies over this time have extracted functional information from angiography based on contrast concentration-time curves, either by indicator dilution or transit-time methods.^34–40 Cerebral aneurysms are regions with distinct flow patterns and can be conveniently demarcated from the vasculature. Therefore, angiographic techniques have been extensively used to quantify intraneurysmal hemodynamics. Basic parameters of aneurysmal concentration-time curves (aC-T curves) such as slopes or mean-transit-time calculations, calculations of variations in image intensity throughout the aneurysm or at the inflow and outflow zones, or curve-fitting of gamma-variate, single- and double-exponential, polynomial, or lagged-normal models to aC-T curves have been used to derive quantities representative of intraneurysmal flow.^{31,32,41–49} Derived quantities are then employed to quantify device efficacy by considering the changes in the quantities pre- and post-treatment.

Based on the observation that the spatial dispersion of dye in flow through tubes was nearly Gaussian and that mixing in cardiac chambers and skewness of measured concentration-time curves seemed to follow an exponential trend, the lagged-normal function was suggested ⁵⁰ as an appropriate equation for fitting concentration-time curves. This function, which is a convolution of a Gaussian function and an exponential decay, can be used to represent the sequential or simultaneous effect of dispersion and mixing processes. This function has previously been used to quantify flow-diverter performance in vitro as well as in vivo.^15,18,41,42

In this in vitro study, we use basic parameters of angiographic aC-T curves along with a lagged-normal-derived function to evaluate five neurovascular devices (three high-porosity stents and two low-porosity flow diverters) in idealized sidewall and bifurcation aneurysm models. While previous in vitro studies^21,51,52 have assessed these devices, a collective, side-by-side comparison of the angiographic performance of these five devices both in sidewall and bifurcation geometries has not yet been conducted.

Methods

A total of five devices were tested. Three of the devices are considered high-porosity stents: Neuroform (Stryker Neurovascular, Fremont, CA), Enterprise (Codman Neuro, Codman & Shurtleff, J&J, Raynham, MA), and Low-Profile Visualized Intraluminal Support (LVIS, MicroVention, Terumo, Tustin, CA). Two of the tested devices are low-porosity stents or flow diverters: Pipeline (Covidien, Medtronic, Irvine, CA) and Flow-Redirection Endoluminal Device (FRED, MicroVention, Tustin, CA). Device dimensions were matched as closely to each other as possible at the time of testing, and were as follows: sidewall model—Neuroform 4 × 20 mm, Enterprise 4.5 × 28 mm, LVIS 4.5 × 34 mm, FRED 4 × (23/17) mm, and Pipeline 4.25 × 20 mm; bifurcation model—Neuroform 4.5 × 20 mm, Enterprise 4.5 × 28 mm, LVIS 4.5 × 34 mm, FRED 4 × (23/17) mm, and Pipeline 3.75 × 20 mm.

Two idealized silicone aneurysm model configurations were manufactured and employed in device testing. The sidewall model was composed of a straight, 4 mm diameter vessel with a 7 mm aneurysm projected perpendicularly from its midportion. The bifurcation model was composed of a 4 mm diameter vessel segment that bifurcated into two 2.5 mm daughter vessels, opposed and 130 degrees from parallel; a midline 7 mm aneurysm projected from the bifurcation (Figure 1).

Figure 1.

Schematic of the sidewall and bifurcation model geometries. The gray shaded region indicates stent deployment location.

Aneurysm models were three-dimensional (3D)-printed and dip-coated in silicone, as described elsewhere in detail. ⁵³ Briefly, the aneurysm model lumen was designed (Solidworks, Dassault Systemes, Waltham, MA) and 3D printed in acrylonitrile butadiene styrene (ABS) via Fused Deposition Modeling (Dimension Elite, Stratasys, Eden Prairie, MN). This 3D printing method leaves print lines resulting in surface roughness of the ABS mold. Prior to casting with silicone, the ABS mold surface was smoothed out by shallow melting the outer layers through repeated dipping in Xylene. The smoothed and processed solid ABS luminal mold was then repetitively dip-coated in silicone to ensure a consistent, bubble-free, and transparent model. After curing for up to 48 h, the ABS luminal core was destroyed to produce the silicone replica.

Under fluoroscopic guidance and the assistance of a neurointerventionalist, the devices were deployed in separate sidewall and bifurcation models. Devices were implanted across the aneurysm neck in sidewall models, and in the case of bifurcation models, a single device was deployed across the aneurysm neck and into one of two daughter vessels (Figure 2). Sidewall and bifurcation replicas without implanted devices served as untreated controls. Deployments were performed in blood analog fluid warmed to 37℃.

Figure 2.

Radiograms of the five devices deployed in the sidewall (top row) and bifurcation (bottom row) models. From left to right: Neuroform, Enterprise, Low-Profile Visualized Intraluminal Support (LVIS), Flow-Redirection Endoluminal Device (FRED), Pipeline.

A flow rig was designed for repeatability during high-speed angiographic acquisition (Figure 3). The setup allowed for the continuous circulation of warmed blood analog during contrast injection proximal to the idealized aneurysm model. A distal starling resistor of 80 mmHg was used to pressurize the entire system to near-physiological levels. This mitigated the effect of static-fluid head and ensured that the complex dynamics of contrast injections were duplicated.

Figure 3.

Schematic of the experimental flow rig.

Flow data were acquired using an in-line (cannulated) flow transducer (ME6PXN, Transonic, Ithaca, NY) and PowerLab data acquisition system (ADInstruments, Colorado Springs, CO). A peristaltic pump was used to drive the fluid through the apparatus with a mean flow rate of approximately 4 ml/s. The flow in the system with this pump and the starling resistor in the series was pulsatile, but the pulsatility was not physiological; the flow waveforms were sinusoidal with an amplitude about three times the mean flow rate at a frequency of about 4 Hz. A 50% glycerol aqueous solution (by mass) served as blood analog fluid with a viscosity of 3.5 cP at 37℃. The mean Reynolds number (fluid density × mean velocity × vessel diameter/fluid viscosity) was 360, which is similar to the mean Reynolds numbers in human cerebrovascular flow. The glycerol solution was stored in a warming reservoir. A starling resistor (set at 200 mmHg) was assembled in parallel to the test section to serve as a pressure relief/bypass valve.

Ultravist 370 contrast, with a specific gravity adjusted to be similar to that of the working fluid (50% contrast/saline mixture by volume), was injected directly into the flow circuit via high-pressure tubing. The injection site was 50 cm proximal to the aneurysm model (approximately 80 tube diameters), allowing sufficient mixing prior to the test segment. Injections were performed using a MedRad power injector (MedRad Mark V ProVis, Bayer Healthcare, Whippany, NJ), at a rate of 2.5 ml/s for 2 s with a 2-second injection delay. Six trials were conducted for each aneurysm model/device combination.

High-speed digitally subtracted angiograms (Artis Zeego, Siemens, Erlangen, Germany) were acquired at 15 frames/s for approximately 20 s during injections to capture contrast wash-in/wash-out characteristics of the aneurysm. The angiograms were transferred in digital imaging and communications in medicine (DICOM) format to MATLAB (MATLAB 2015, MathWorks, Natick, MA) for analysis.

From the acquired image sequence, a region of interest (ROI) containing the aneurysm was manually segmented (roipoly function in MATLAB) and the grayscale intensity within the ROI was quantified to obtain aC-T curves (Figure 4). The curves were then quantified with six parameters: convective decay constant (τ_conv, s), diffusive decay constant (τ_diff, s), time to peak (TTP, s), mean transit time (MTT, s), full-width-at-half-maximum (FWHM, s), and washout slope (WOS, s^–1).

Figure 4.

Averages (n = 6 each) of the normalized aneurysmal concentration-time curves for the sidewall (a) and bifurcation (b) models. C: control; N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline.

The convective and diffusive decay constants were derived by least-squares fitting of the aC-T curves with a mathematical model developed previously (Figure 5).^{17,18,35,41,42} The model has been described in detail previously, ⁴¹ but briefly it consists of lagged-normal, cumulative, and exponential decay functions that demarcate the contrast transport represented by the aC-T curves to convective and diffusive modes of transport. The lagged-normal component represents convective transport of contrast that washes in and out of the aneurysm (nominally, closer to the aneurysm neck) and the convective decay constant quantifies this convective delay. The cumulative function represents contrast that enters the aneurysm but remains entrapped in recirculation or stagnant regions (nominally, closer to the aneurysm dome), while the exponential decay function characterized by the diffusive decay constant represents the contrast exiting the aneurysm by diffusion from these stagnant regions. Thus, a larger proportion of the aneurysmal contrast transport is governed by the diffusive component in devices with better flow-diversion efficacy, or to put it differently, the diffusive mode of transport has a greater amplitude in flow diverters as compared to stents.

Figure 5.

A representative mathematical model-fit to an aneurysmal concentration-time curve.

TTP is the time taken to reach maximum contrast concentration in the aneurysm. The MTT represents the centroid of the aC-T curve. The full-width-at-half-maximum is the width of the concentration-time curve at half of the peak contrast concentration. The washout slope (WOS, s^–1) is the average slope of the aC-T curve from the point of peak concentration till the concentration at the end of the acquisition period. If the contrast concentration decayed to 2% or less of the peak concentration within the acquisition period, the WOS was calculated as the average slope of the aC-T curve from the point of peak concentration to the point at 2% of peak concentration. This ensured that the WOS of aC-T curves with rapid decays was adequately captured. TTP, MTT, and FWHM are commonly used parameters to quantify concentration-time curves, as is the WOS parameter, which has previously ³² been used for angiographic quantification of flow-diverter performance. While all four parameters represent intraneurysmal contrast delay or stagnation, they are influenced by slightly different phenomena. During contrast injection, the pressure in the vasculature is increased and thus the contrast is, in some sense, “forced into” the aneurysm. Thus, while the device permeability is reflected in the TTP measure, it can also be heavily influenced by the injection parameters. The WOS parameter calculates the slope after injection has stopped and is thus a basic parameter that reflects only the contrast wash-out phase. The MTT and FWHM parameters are influenced by both the injection/wash-in and wash-out phase. Stent or flow-diverter implantation across an aneurysm can be expected to increase τ_conv, τ_diff, MTT, TTP, and FWHM and decrease WOS.

Statistical comparison of the parameters across the six cases (control and five devices) was performed using GraphPad InStat (GraphPad, La Jolla, CA). Comparison between the groups was performed using one-way analysis of variance (ANOVA) test for parametric or normally distributed data. For non-normal data sets, the Kruskal–Wallis test was used.

Results

Representative aC-T curves for the control and five device deployments within sidewall and bifurcation aneurysm models yielded two basic profiles (Figure 4). As can be expected, control and high-porosity devices (Neuroform, Enterprise, and LVIS) demonstrated a more rapid TTP and contrast wash-out in the sidewall model. Lower-porosity devices (FRED and Pipeline) demonstrate more gradual aneurysm filling and subsequent contrast stagnation. Only four of the six LVIS sidewall trials showed a diffusive component to the contrast transport. None of the other devices or controls showed any diffusive component in either the sidewall or the bifurcation models (LVIS bifurcation had no diffusive component either). The diffusive decay constant (τ_diff) was thus discarded from the device comparisons. Although there were some statistically significant differences between the devices in the bifurcation models, in general all the devices performed similarly. The mean and standard deviation of the five parameters quantified from the aC-T curves (MTT, TTP, FWHM, WOS, and τ_conv) for the control and five device groups are presented in Table 1. Figures 6–9 show inter-device comparisons of τ_conv, MTT, TTP, and WOS, respectively; the plots are normalized by the corresponding average control values.

Table 1.

Mean ± standard deviation of all the angiographic parameters evaluated for all the devices. Statistical significance is noted for devices compared to the control (no device) case.

	Parameter	Control	Neuroform	Enterprise	LVIS	FRED	Pipeline
SIDEWALL	MTT (s)	5.45 ± 0.08	5.82 ± 0.37	5.11 ± 0.08	6.54 ± 0.69d	8.02 ± 0.30d	9.32 ± 0.29d
	TTP (s)	3.26 ± 0.05	3.47 ± 0.17	3.19 ± 0.05	3.21 ± 0.23	4.20 ± 0.25c	5.76 ± 0.87d
	FWHM (s)	4.94 ± 0.13	5.81 ± 0.53	4.64 ± 0.10	5.75 ± 0.24	9.95 ± 1.29d	15.29 ± 1.02 e
	WOS (s–1)	0.07 ± 0.002	0.07 ± 0.009	0.08 ± 0.005	0.06 ± 0.007 b	0.05 ± 0.003d	0.04 ± 0.004d
	τconv (s)	3.80 ± 0.09	3.98 ± 0.60	3.23 ± 0.09	3.70 ± 0.41	10.12 ± 1.65d	18.35 ± 3.31d
BIFURCATION	MTT (s)	2.96 ± 0.32	3.62 ± 0.06d	3.42 ± 0.10c	3.28 ± 0.16	3.55 ± 0.15d	3.70 ± 0.19d
	TTP (s) a	2.27 ± 0.10	2.05 ± 0.08	2.18 ± 0.15	2.37 ± 0.05	2.17 ± 0.12	2.15 ± 0.18
	FWHM (s)	2.60 ± 0.05	2.61 ± 0.04	2.69 ± 0.07	2.55 ± 0.07	2.67 ± 0.08	2.94 ± 0.33c
	WOS (s–1) a	0.14 ± 0.052	0.07 ± 0.009	0.07 ± 0.014	0.08 ± 0.025	0.07 ± 0.009c	0.07 ± 0.013
	τconv (s) a	0.67 ± 0.08	1.61 ± 0.05c	1.23 ± 0.33	0.74 ± 0.05	1.19 ± 0.30	1.81 ± 0.05d

^aNonparametric analysis of variance.

^bp < 0.05, ^cp < 0.01, ^dp < 0.001.

^eSidewall Pipeline FWHM could not be calculated for most trials because the contrast intensity did not reduce to half-peak value in the washout phase (high-contrast residence time); Pipeline not included in statistical comparison.

MTT: mean transit time, TTP: time to peak, FWHM: full width at half maximum; WOS: washout slope; τ_conv: convective decay time constant; LVIS: Low-Profile Visualized Intraluminal Support; FRED: Flow-Redirection Endoluminal Device.

Figure 6.

Convective time constant for the five devices as a percentage of the average control (no device) case for the sidewall and bifurcation models. τ_conv: convective decay time constant; N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline, n = 6 each, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 7.

Mean transit time (MTT) of the five devices as a percentage of the control (no device) average for the sidewall and bifurcation models. N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline, n = 6 each, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 8.

Time-to-peak (TTP) of the five devices as a percentage of the control (no device) average for the sidewall and bifurcation models. N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline, n = 6 each, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 9.

Washout slope (WOS) of the five devices as a percentage of the control (no device) average for the sidewall and bifurcation models. N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline, n = 6 each, *p < 0.05, **p < 0.01, ***p < 0.001.

Sidewall model

The FRED and Pipeline devices showed statistically significantly better flow diversion as compared to LVIS, Neuroform, and Enterprise devices in all the inter-device parameter comparisons (there was a single exception in which the FRED WOS was better than the LVIS WOS, but without statistical significance). The Pipeline device resulted in a substantially delayed washout, due to which the FWHM parameter could not be calculated for most Pipeline trials (the contrast washout phase did not decay back down to the half-maximum value). The Pipeline device performed significantly better than the FRED device in all parameter comparisons. The three stent devices were statistically equivalent in the τ_conv and TTP comparisons. The LVIS device was statistically better than both the Neuroform and Enterprise devices in the MTT and WOS comparisons. The Neuroform and Enterprise devices were essentially statistically equivalent to each other and to the control case (Table 1).

Bifurcation model

Device performance was scattered in the bifurcation model with no clear statistical conclusions, but some general trends can be noted. Overall, the Pipeline device performed marginally better than all other devices. The Pipeline device was better than the FRED device in the τ_conv and FWHM parameters (statistically significant) as well as the MTT (not significant), but was worse than the FRED in the TTP and WOS parameters (not significant). The TTP was the best with the LVIS device as compared to all other devices (statistically better than Neuroform only), but the LVIS performed the worst in all other parameter comparisons. The TTP was the worst with the Neuroform device as compared to all other devices. The Neuroform device was better than both the Enterprise and LVIS stents in the MTT, WOS, and τ_conv parameters; the Neuroform FWHM was worse than the Enterprise device (not statistically significant). The Neuroform convective decay constant (τ_conv) was statistically better than even the FRED device.

Discussion

To the knowledge of the authors, this is the first in vitro study comparing the intra-aneurysmal hemodynamics within sidewall and bifurcation aneurysm models of these five neurovascular devices (previous in vitro studies have compared some of these devices individually). The primary limitation of the study is that device performance comparisons are made on single deployments owing to cost and availability of devices. Repeated deployments would allow for more robust comparisons. Similarly, we compared the devices at only one flow rate (∼4 cc/s). Additional flow rates may have provided a better picture of inter-device performance, but the primary parameter that affects device performance is the net permeability of the devices, and device permeability is independent of flow rates. The experimental setup we used helped mitigate variability in the inter-device aC-T curves, but the variability was greater than expected owing, possibly, to the timing of injections to the peristaltic pump cycle or fluctuations from the starling resistor. A larger number of trials per device might have improved some of the statistical results, but as can be seen in Figures 6–9, the median coefficient of variation (standard deviation/mean) over all the parameters compared was only 6% (sidewall range 1%–18%, n = 29; bifurcation range 1%–38%, n = 30) and justifies the overall conclusions drawn below.

It should be noted that the angiographic quantification used here essentially tracks the transport of contrast medium into and out of the aneurysm. If the injection is performed properly so that the contrast homogenously mixes with blood, the transport of contrast will represent the transport of blood through the aneurysm. ⁵⁴ But this methodology does not provide a detailed assessment or quantification of blood flow velocities within the aneurysm like particle image velocimetry (PIV) or CFD studies. Local velocity patterns or wall shear stress measures within the aneurysm can only be quantified by PIV/CFD studies although some localized parameters or flow structures have previously been extracted by angiographic studies.^31,32 It is also not yet clear whether detailed velocity measures from PIV/CFD studies are necessary to evaluate the overall performance of flow diverters or to predict flow-diversion treatment outcome in patients. A previous summary ¹⁰ of CFD and PIV studies suggests that lower-porosity flow diverters and/or sidewall geometries can induce ∼>90% reduction in intraneurysmal velocity measures as compared to respective controls, whereas higher-porosity stents and/or bifurcation geometries can result in ∼<40% reduction in intraneurysmal velocity. The angiographic quantification used here results in parameter values that are very much in line with these CFD and PIV velocity reductions. An overview of Figures 6–9 shows that the stents caused minimal change in intraneurysmal flow activity as compared to the control both in the sidewall and bifurcation geometries; the flow diverters also have similar flow activity as control in the bifurcation geometry but cause substantial reduction in flow activity in the sidewall geometry.

We did not use a catheter for the injections because power injections through a catheter can result in artifacts in the aC-T curves. The large fluid capacity of the power injector syringe combined with the relatively rapid injection rates causes the contrast to exit the catheter tip even after completion of injection due to inertia. Contrast is swept out of the catheter tip during each systolic pulsation and can result in small boluses entering the aneurysm every cycle resulting in oscillatory artifacts in the aC-T curves. We thus chose a direct injection into the flow circuit to mitigate these artifacts. The site of injection was also 50 cm proximal to the aneurysm to allow for complete mixing of the contrast with working fluid (blood). While this is a supra-physiological distance, a previous study suggests that mixing of contrast with blood may be complete within approximately 10 arterial diameters. ⁵⁴ Thus, in patients, where aneurysm parent vessel diameters are about 3–5 mm, an injection in the proximal/extracranial carotid or vertebral arteries should more than suffice to allow for contrast mixing.

The changes in intra-aneurysmal flow characteristics demonstrated by flow diversion for the sidewall aneurysm model were not reflected in the bifurcation aneurysm models. As noted above, the angiographic quantification used here may not have been sensitive enough to detect any inter-device differences while detailed velocity measures provided by PIV or CFD studies might have shown statistically significant differences. Although statistically conclusive statements cannot be drawn from the angiographic bifurcation results, a couple of observations can be made based on the evidence. First, the Pipeline device performed better than the other devices, which can generally be expected because of its design characteristics as a flow diverter. We did not quantify wall apposition in the study, but the radiograms suggest slightly poor apposition of the Pipeline deployment to the inner curvature of the bifurcation geometry as compared to the other devices (Figure 2, bottom row). A corresponding compression of the wires to the outer curvature facing the aneurysm neck might have contributed to its improved performance (especially as compared to the other flow diverter). Second, the Neuroform device seemed to perform the second best after the Pipeline in the bifurcation geometry. The better apposition of the Neuroform’s open-cell design to the aneurysm neck (Figure 2, bottom row) might have contributed to this effect. Better mesh apposition to the neck even with minimal to low metal coverage structures might thus provide equivalent flow diversion as high metal-coverage meshes that are displaced from the aneurysm neck. Flow diverters have been shown to fail to occlude experimental bifurcation aneurysms while occluding sidewall aneurysms in a canine model. ⁵⁵ As noted above, experimental and numerical studies have shown a spectrum of flow-diversion response with low to minimal effects of the device on bifurcation-type geometries as compared to sidewall-type geometries. ¹⁰ Our results reflect these studies and underscore the differences in the effectiveness of flow diversion depending on aneurysm geometry.

Three general conclusions can be drawn based on the results in the sidewall model:

• 1. The lower-porosity (and higher-pore density) flow-diverter devices performed significantly better in reducing intraneurysmal flow activity than the higher-porosity stents. This effect is expected, has been studied extensively, and is well established.

• 2. Generally, the evidence from this specific set of experiments suggests that the LVIS device performed better than the Neuroform or Enterprise devices. While the Neuroform and Enterprise devices were equivalent to the control case in all measured parameters (Table 1), the LVIS was statistically better in two of the parameters. Also, our mathematical model indicates a diffusive component to the aC-T curves when the tail end of the washout phase “flattens” out, indicating increased contrast residence time. This diffusive time constant was not quantified here because the LVIS was the only device that showed a diffusive component; the inclusion of that parameter would have weighted the LVIS toward superior flow-diversion characteristics (in comparison to the Neuroform and Enterprise, at the least). A previous experimental study ⁵¹ found that the flow diversion from a braided intermediate- (88%) porosity stent was better than high- (∼93%) porosity stents and similar to those of flow diverters (∼70% porosity). To the extent that the porosity of the LVIS device is lower than “standard” neurovascular stents, its flow-diversion behavior is potentially better. A CFD study ⁵⁶ on three patient-specific geometries (sidewall-type geometries on the outer curvature) also found the flow-diversion performance of the LVIS stent to be in between that of the Enterprise and Pipeline devices.

• 3. The Pipeline device performed statistically better than the FRED in all five parameters considered. The FRED device is supposed to have a lower nominal porosity and higher nominal pore density than the Pipeline device, and as such this result is surprising. To mitigate any effects from the pressurized system with starling resistors, we ran a separate short study comparing the same FRED deployment to a new Pipeline deployment with a centrifugal pump open to atmospheric pressure (flow rate and blood analog fluid were kept the same as the main study). The results confirmed the “weaker” performance of the FRED as compared to the Pipeline (MTT FRED = 121 ± 13% of control average, MTT Pipeline = 139 ± 4% of control average, p < 0.01, n = 7). A recent in vitro study compared aneurysmal flow alteration between five flow diverters including the FRED and Pipeline based on the reduction in dynamic pressure. ⁵² That study also found the Pipeline to fare better than the FRED. The importance of cell geometry to device permeability ⁴⁸ and fluctuations in thin wire structures causing increased intraneurysmal flow ¹³ have been evaluated previously. It is possible that some parameter other than just porosity and pore density affects net device permeability, or the unique double-layer construction of the FRED potentially mitigates performance. The Pipeline used in our study was a 4.25 mm device, while the FRED was a 4.0 mm device deployed into a replica of 4 mm nominal diameter (the actual replica diameter may have potentially reduced by 50–100 microns during the manufacturing process). Oversizing can cause changes in the net porosity and permeability, ⁵⁷ but this is an unlikely cause of the observed effect sizes. A CFD study based on micro-computed tomography (CT) scans of in vitro deployments showed the Pipeline and FRED to induce equivalent reductions in intraneurysmal velocity. ¹⁰ At any rate, further experiments with multiple deployments must be evaluated to confirm or reject this finding.

The porosity and pore density of these devices have been measured previously ¹⁰ based on micro-CT scans. Figure 10 shows a plot of the convective time constant against device porosity and pore density for the sidewall aneurysms. Correlations of MTT versus device porosity (R²= 0.57, p = 0.14) and pore density (R²= 0.66, p = 0.09) show similar trends. The correlations are not statistically significant, but the trends show that lower device porosity or higher device pore density can lead to reduced intraneurysmal flow activity post-treatment (increase in MTT or convective time constant). The sample sizes are low, but again, the lack of statistical significance may be because some parameter other than porosity and pore density is required to fully characterize the hemodynamics after flow diversion. It may be useful to evaluate the relationship between angiographic parameters such as those used in our study and detailed hemodynamic parameters such as hydrodynamic circulation and kinetic energy obtained by CFD simulations or PIV experiments. The convective time constant from this study was plotted against the mean intra-aneurysmal kinetic energy from a previous CFD study ¹⁰ on the same devices in a sidewall aneurysm (Figure 11). As expected, the comparison shows an inverse relation between the kinetic energy and the convection time constant (R²= 0.52, not statistically significant). These two parameters, convective time constant and kinetic energy, can potentially serve as predictive indices of flow-diverter efficacy. Larger data sets comparing angiographic to detailed hemodynamics parameters are needed to reliably increase the clinical utility of angiographic quantification, which can be performed in real-time as each patient is being treated. The effect of patient-to-patient variation in coagulation profiles and/or the formation of a neointimal layer must also be considered when evaluating flow-diversion outcomes.

Figure 10.

Average intraneurysmal convective decay time constant (percentage of the control) plotted against device porosity and pore density for the sidewall aneurysm. N: Neuroform; E: Enterprise; L: Low-Profile Visualized Intraluminal Support (LVIS); F: Flow-Redirection Endoluminal Device (FRED); P: Pipeline; porosity and pore density obtained from Dholakia et al. ¹⁰ The trends (porosity: solid line, pore density: dashed line) are not statistically significant.

Figure 11.

Intraneurysmal mean kinetic energy (percentage of the control) obtained from computational fluid dynamics plotted against the average intraneurysmal convective decay time constant (percentage of control) obtained from angiography. The trend is not statistically significant.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: RJD is currently an employee of MicroVention. HHW receives royalties from Codman. DJF has consulted for Medtronic, Stryker, and MicroVention, and receives royalties from Codman. All other authors have nothing to declare.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.