Volume-based sizing of the Woven EndoBridge (WEB) device: A preliminary assessment of a novel method for device size selection

Abstract

Introduction

Accurate sizing of the Woven EndoBridge (WEB) device is of critical importance as it determines procedural safety and successful occlusion of wide neck bifurcation aneurysms. The aim of this study was to assess the ability of aneurysm volume to assist in accurate WEB size selection.

Methods

All patients with an intracranial aneurysm treated with the WEB SL or WEB SLS device between March 2019 and October 2019 were identified for this retrospective study. Aneurysm volumes were calculated with auto-segmentation using a three-dimensional volume rendering program on an independent Syngo workstation (Siemens Healthineers AG). Pearson correlation coefficients were calculated for aneurysm auto-segmented volumes and WEB volumes, as well as for aneurysm height × width and WEB height × width. Follow-up angiographic outcomes were collected at 6–9 months post-procedure.

Results

Twenty-nine aneurysms were evaluated by 3D rotational angiography. The correlation coefficient with WEB size was larger for auto-segmented aneurysm volumes (r = 0.979) compared to height × width measurements (r = 0.867). Using Fisher r-to-z transformations, we found the difference between the two correlations to be statistically significant (p = 0.0007). Follow-up angiography available in 13 subjects demonstrated an 85% complete aneurysm occlusion rate.

Conclusion

Aneurysm volumes are highly correlated with WEB volumes, with auto-segmentation volumes displaying statistically significant difference against conventional height by width measurements. These results suggest that volumetric measurements of aneurysm size provide a useful adjuvant measure to assist in appropriate size selection of the WEB device.

Introduction

Endovascular embolization has quickly evolved to become an effective first line treatment for ruptured and unruptured intracranial aneurysms at many institutions.^1,2 However, wide-neck bifurcation aneurysms present a challenge to endovascular management and often require adjunctive techniques such as balloon-remodeling or stent-assisted coiling.^3,4 While numerous endovascular strategies have been developed to address these difficult lesions, none are perfectly effective. The FDA recently approved the Woven EndoBridge (WEB) (Microvention, Aliso Viejo, California) device as an intrasaccular flow diverter for wide neck bifurcation aneurysms. ⁵ The device is designed to provide flow disruption at the aneurysm ostium leading to gradual thrombosis and obliteration of the aneurysm.^6,7 It has evolved from the initial double layered version (WEB DL) to the now single layered versions (WEB SL and WEB SLS).

Previous registries have shown good safety results (WEB-IT, WEBCAST, WEBCAST 2, and the French Observatory) with the use of the WEB device.^8–12 The primary risks to consider during deployment are either oversizing the device leading to thromboembolic events, or undersizing the device leading to poor occlusion rates. Successful intrasaccular flow diversion depends on appropriate sizing. The effectiveness of the device is determined by its apposition to the aneurysm wall which generates increased metal coverage over both the neck and dome. ¹³ Furthermore, inappropriate sizing not only increases procedural risks but also causes the angiographer to open multiple devices for a single case leading to waste and increased cost. ¹⁴ Accurate device sizing is important for patient safety, optimal outcomes, and decreased cost.

WEB size is traditionally calculated by averaging multiple height and width measurements of the aneurysm dome. Because height and width measurements incompletely describe the three-dimensional morphology of an aneurysm, this method may be misleading when selecting an appropriately sized intra-saccular flow diverter and leads to a trial-and-error type of learning process and a steep learning curve for clinicians. During the WEB-IT trial, device size was determined by using a neurovascular simulator (Vascular Simulations) ¹⁵ ; however, this is not always readily available to all practitioners. Therefore, a methodology that can enhance WEB size based on three-dimensional aneurysm morphology would be extremely valuable. The goal of this study was to assess the correlation of auto-segmented aneurysm volumes with implanted WEB size.

Methods

This is a single center study approved by the local institutional review board. We performed a retrospective study of 29 consecutive patients with at least one ruptured or unruptured intracranial aneurysm treated with the WEB SL or WEB SLS device between March 2019 and October 2019. All patients were treated under general anesthesia using flat panel biplane angiography (Artis Q, Siemens Healthineers AG, Erlanger, Germany). WEB devices were delivered with tri-axial access using either the VIA 21, VIA 27, or VIA 33 microcatheter (Microvention, Aliso Viejo, California). The recorded data included the location of aneurysm, aneurysm rupture status, utilization of a stent device, and the total number of WEB devices used per procedure. Analysis of aneurysm anatomy was performed as described below. Follow-up angiographic outcome was assessed with digital subtraction angiography (DSA). Aneurysm occlusion was assessed using the WEB Occlusion Score (WOS). ¹⁶

All aneurysms were evaluated using a three-dimensional volume rendering software on an independent Syngo workstation (Siemens Healthineers AG) prior to treatment. Anatomical data (maximal dome height, width, depth, and neck size) were measured manually for each aneurysm, in addition to dome height-to-neck ratios. By convention, WEB device size is selected based on aneurysm height and width measurements in two orthogonal planes. Therefore, aneurysm heights on two orthogonal planes was determined to be the “height” and the average of aneurysm width and depth was determined to be “width” in linear measurement analyses. Next, these linear measurements were used to derive ellipsoid volume ([π × height × width × depth]/6) and cylindrical volumes (π × radius² × height), which closely approximate aneurysm volumes. Finally, aneurysm volume data were derived from automated segmentation software with thresholding on the Syngo workstation. Thresholds were set manually by placing seed pixels within regions that designated the proximal parent vessel, aneurysm dome, and distal parent vessel. Implanted WEB devices were assessed for best fit by the treating physician with high resolution contrast enhanced cone beam CT (DynaCTA) before detachment of the device.

Baseline variables were described using descriptive statistics. Continuous variables were reported as median and interquartile range. Categorical variables were reported as frequency. Pearson correlation coefficients were calculated for ellipsoid, cylindrical, and aneurysm auto-segmented volumes and WEB volumes (obtained from manufacturer), as well as for aneurysm height × width and WEB height × width. Linear regression analysis was completed for each independent variable: Auto-segmentation volume, cylindrical volume, ellipsoid volume, and the product of aneurysm height × width. Fisher r-to-z transformations were used to assess the significance of the difference between two correlation coefficients. The level of statistical significance was set at 0.05. Additionally, the percent difference between WEB size and aneurysm size was calculated for auto-segmented volume-based measurements. Differences greater than or less than 20% were classified as oversized or undersized WEB device placement.

Results

Twenty-nine aneurysms were treated with the WEB device. The anatomical data is listed in Table 1. The median aneurysm height and width were 4.3 mm and 4.1 mm, respectively. Median dome-to-neck ratio was 1.2. By volume, the largest aneurysm was 916.4 mm³, and the smallest was 14.0 mm³, with a median volume of 84.7 mm³. The smallest WEB device used was 4×2.6 and the largest WEB device was 11×8. Twelve anterior communicating artery (ACOM) aneurysms, 6 middle cerebral artery bifurcation (MCA) aneurysms, 5 basilar apex (BA) aneurysms, 4 posterior communicating artery (PCOM) aneurysms, 1 internal cerebral artery terminus (ICAT) aneurysm, and 1 superior hypophyseal artery (HYP) aneurysm were treated. Six (21%) aneurysms were ruptured at the time of treatment. Seven (24%) aneurysms required the use of the WEB SLS device, while the remaining 22 (76%) were treated with the WEB SL device. The WEB 17 system was not used due to lack of availability. Five (17%) aneurysms were treated with the WEB device in conjunction with the LVIS (Microvention, Aliso Viejo, California) stent or Pipeline Flex (Medtronic, Minneapolis, Minnesota) device within the parent vessel. Some procedures required re-sizing of the WEB device after deployment. Specifically, 35 WEB devices were used for 29 procedures, resulting in a WEB devices-per-procedure ratio of 1.2. Five devices were removed due to either parent vessel encroachment or residual aneurysm filling, and 1 device was removed due to inadequate shape conformability. Considering final implanted WEB devices, 6 (21%) and 10 (34%) devices were undersized and oversized, respectively, based on aneurysm volume, however these devices were determined to be adequately sealing the aneurysm ostium on DynaCTA and were not replaced (Table 2). At the time of writing, 45% (13/29) of patients have undergone angiographic follow-up. At a median follow-up of 7 months, 85% (11/13) of aneurysms were completely occluded or demonstrated opacification of only the marker recess (WOS Grade A or B). The remaining 15% (2/13) aneurysms demonstrated residual filling (WOS Grade D).

Table 1.

Aneurysm anatomical data ordered by aneurysm volume.

Aneurysm ID	Aneurysm location	Height (mm)	Width (mm)	Depth (mm)	Aneurysm volume (mm3)	Implanted WEB	WEB volume (mm3)
1	ACOM	3.8	2.6	2.5	14.0	4×2.6	30
2	BA	2.1	3.2	3.3	22.5	4×3	38
3	ACOM	2.0	2.8	3.8	25.7	4×3	38
4	PCOM	2.6	3.8	3.4	31.4	4×3	38
5	ACOM	5.3	4.0	3.0	37.7	5×3.6	70
6	MCA	3.0	3.3	3.9	41.5	4×3	38
7	ACOM	5.2	2.7	4.0	50.2	5×3.6	70
8	ICAT	4.2	3.9	3.8	58.3	5×3.6	70
9	BA	6.4	7.1	6.8	61.1	5×4	79
10	BA	5.3	5.8	4.4	63.0	5×4	79
11	ACOM	2.0	5.2	5.5	63.4	4×3	38
12	ACOM	4.1	3.2	3.5	69.3	5×3	59
13	MCA	4.2	2.5	3.5	80.0	5×4	79
14	BA	3.5	3.7	4.5	81.3	6×3	85
15	ACOM	5.0	3.4	4.0	84.7	6×4.6	130
16	BA	3.4	5.3	5.4	86.2	6×3	85
17	ACOM	3.8	4.6	4.8	88.0	6×3	85
18	MCA	3.1	5.0	6.6	88.1	6×3	85
19	ACOM	4.4	2.5	3.0	92.7	5×3.6	70
20	ACOM	3.3	4.9	5.2	99.5	5×3	59
21	PCOM	7.9	3.4	2.8	103.0	4×3	38
22	ACOM	6.8	6.0	3.9	122.3	7×5.6	220
23	PCOM	7.4	6.7	6.4	162.3	7×4	154
24	MCA	5.1	4.5	4.2	168.4	6×4	113
25	MCA	4.3	6.4	6.4	191.3	7×5	192
26	ACOM	5.7	4.1	5.4	246.5	7×5	192
27	MCA	7.6	6.2	3.9	313.9	9×5	318
28	HYP	11.3	9.3	9.1	560.0	10×7	550
29	PCOM	9.5	7.9	7.0	916.4	11×8	760

ACOM, anterior communicating artery, BA, basilar artery, PCOM, posterior communicating artery, MCA, middle cerebral artery, ICAT, internal carotid artery terminus, HYP, superior hypophyseal artery.

Table 2.

Baseline characteristics of data.

	N = 29
Location
ACOM	12	41%
MCA	6	21%
BA	5	17%
PCOM	4	14%
ICAT	1	5%
HYP	1	3%
Ruptured Aneurysm	6	21%
Number of SLS	7	24%
Number of Stent-Assisted	5	17%
Oversized WEB	10	34%
Undersized WEB	6	21%
WEBs Not Implanted	6	21%
	Median	IQR
Height	4.3	3.4–5.7
Width	4.1	3.3–5.8
Depth	3.9	3.5–4.8
Aneurysm Volume	84.7	58.3–122.3
WEB Volume	79.0	59.0–130.0

ACOM, anterior communicating artery, MCA, middle cerebral artery, BA, basilar artery, PCOM, posterior communicating artery, ICAT, internal carotid artery terminus, HYP, superior hypophyseal artery.

All Pearson correlations between each measurement modality and WEB size were statistically significant. The correlation coefficient was largest for auto-segmented aneurysm volumes (r = 0.979, p < 0.0001), followed by height × width (r = 0.867, p < 0.0001) and ellipsoid volumes (r = 0.817, p < 0.0001), and smallest for cylindrical volumes (r = 0.813, p < 0.0001). Pearson correlations coefficients are reported for each association in Table 3. Linear regression models of auto-segmented aneurysm volumes vs. WEB volumes and aneurysm height × width vs. WEB height × width are depicted in Figure 1.

Table 3.

Pearson correlation coefficients and probability values of difference between correlations.

	WEB volume	p
Automated segmentation volume	0.979	0.0007*
Cylindrical volume	0.813	0.5092
Ellipsoid volume	0.817	0.5352
	WEB Height × Width
Height × Width	0.867	N/A

Note: The correlation coefficients of the volume-based measurements with WEB size were tested against the correlation coefficient of height × width using r-to-z transformations. The p-values are reported.

*The difference between the correlation of height × width with WEB size and auto-segmented volume with WEB size was statistically significant.

Figure 1.

Linear regression curves of automated segmentation of aneurysm volume using 3 D volume rendering on Syngo workstation against implanted WEB volumes (p < 0.001, R² = 0.9515) and aneurysm height × width against WEB height × width (p < 0.001, R² = 0.7509). Using Fisher r-to-z transformations, the Pearson correlation coefficients are determined to be significantly different (*p = 0.0007).

The two-tailed probability values of the difference between volume-based correlation coefficients and the height × width correlation is listed in Table 3. Importantly, the difference between auto-segmented volume correlation with WEB size and aneurysm height × width correlation with WEB size is statistically significant (p = 0.0007). Ellipsoid and cylindrical volumes were not significantly different than height × width measurements in sizing the WEB device.

Case ILLUSTRATION

The patient is a 63-year-old female with past medical history of lung cancer who was found to have a left middle cerebral artery aneurysm on MRI for the evaluation of metastatic disease. Digital subtraction angiography demonstrated a 9.6×8.2×6.8 mm left middle cerebral artery bifurcation aneurysm with a wide neck (Figure 2(a)). The patient was placed on aspirin 325 mg in preparation for WEB embolization. The right femoral artery was accessed using an 5 F Fubuki (Asahi Intecc, Tustin, California) guide sheath and a SOFIA EX (Microvention, Aliso Viejo, California) intermediate catheter was navigated into the distal left internal carotid artery. A 3D rotational angiogram of the left internal carotid artery circulation was performed to obtain multiple orthogonal linear measurements of the aneurysm as well as auto-segmented aneurysm volume using an independent Syngo XWP workstation. The height and width were measured to be 9.6 mm and 6.8 mm, respectively (Figure 2(b) to (e)). The auto-segmented volume was calculated to be 313.9 mm³. Using the ‘+1/−1’ rule for sizing the WEB device recommended by most practitioners (described below), the most appropriate device would be an 8×6 mm WEB SL. We determined that a 9×6 mm WEB SL with a device volume of 318.0 mm³ would be sufficiently conformable and achieve better mechanical stability within the aneurysm. The aneurysm was catheterized with a VIA 27 microcatheter over a Synchro2 (Stryker Neurovascular, Fremont, California) microguidewire and a 9×6 mm WEB SL was deployed and detached within the aneurysm sac. Post-procedure DynaCTA demonstrated complete coverage of the aneurysm ostium without parent vessel encroachment and post-procedure angiography demonstrated contrast stagnation within the aneurysm. Follow-up angiography at 6 months showed significant vessel remodeling and complete (WOS Grade B) occlusion of the aneurysm (Figure 2(f)).

Figure 2.

WEB embolization of a left middle cerebral artery bifurcation aneurysm using volumetric analysis. (a) Digital subtraction angiography demonstrating a 9.6×8.2×6.8 mm left middle cerebral artery bifurcation aneurysm. (b) Three-dimensional rotational angiographic reconstruction of the aneurysm. (c) Auto-segmentation of the aneurysm is completed by placing seed pixels (colored markers) within the parent vessel, aneurysm, and branching vessel. (d) Segmentation analysis determined the volume of the aneurysm to be 303.9 mm³. (e) A rotated view of the aneurysm shows the height and width are 9.6 mm and 6.8 mm, respectively. (f) At 6 months follow-up, digital subtraction angiography demonstrates WOS Grade B obliteration of the patient’s aneurysm.

Discussion

Although WEB-based embolization has been shown to be an effective treatment of wide-necked intracranial aneurysms, complication rates are not negligible.^17–19 The largest meta-analysis to date to evaluate the efficacy and safety of the WEB device included 940 patients with 962 intracranial aneurysms. ²⁰ Including all large prospective series, it found overall periprocedural complication rate to be 14% (130/962), the majority of which were thromboembolic (91/130) and only a small number of hemorrhagic events (14/130). WEB placement was technically feasible in 97% of patients leading to adequate aneurysm occlusion in 81% of patients. In a combined series of the French Observatory, WEBCAST, and WEBCAST 2, although the periprocedural complication rate was low (1.2%), device placement was not feasible in 4% of cases due to six failures related to lack of appropriate size availability. ²¹ Moreover, the 12-month outcomes of the WEB-IT trial showed complete aneurysm occlusion rate to be only 53.8%, although adequate occlusion was achieved in 84.6% of aneurysms. ⁹ While occlusion rates of challenging wide-neck bifurcation aneurysms are significantly better with WEB-embolization compared to conventional endovascular methods,^22,23 there is an obvious need to maximize the efficacy of the WEB device.

Accurate sizing of the WEB device is of critical importance. Periprocedural complications and inadequate long-term occlusion can be directly related to inappropriate device size. Large devices may protrude into the parent vessel and result in thromboembolic complications, whereas undersized devices incompletely appose the aneurysm wall and poorly cover the aneurysm ostium which leads to incomplete long-term occlusion. ²⁴ In the WEB-IT trial, WEB choice was pre-planned based on neurovascular simulation. ¹⁵ Simulator-based training may increase procedural safety but may not be readily available to all practitioners. Therefore, we sought to identify a methodology to select WEB size that obviates the need for neurovascular simulation.

Typically, size selection of the WEB device is based on exact calibrated measurements of the fundus height and neck width on 3D rotational angiography. In particular, the height and width are calculated as the average of 2 or 3 measurements, and the neck is measured on two orthogonal planes. ⁷ Using height and width measurements, devices are slightly oversized in order to achieve stable conformations within the aneurysm. This has led to the development of the so-called ‘+1/−1 rule’ in which 1 mm is added to the width of the device, and 1 mm is subtracted from the height to allow the device to radially compress against the aneurysm wall and expand longitudinally when deployed. This allows better apposition of the device against the aneurysm wall, greater coverage of the aneurysm ostium, and, importantly, improves occlusion rates by preventing compression of the device over time.^25–27 Herbreteau et al. evaluated the effect of WEB shape modification on the rate of aneurysmal occlusion over time in a small cohort of patients. ²⁸ They found that for devices that were slightly oversized at the time of placement, adequate occlusion was observed in 26/28 (92.9%) aneurysms, whereas only 7/10 (70.0%) aneurysms achieved adequate occlusion with undersized devices. This emphasizes the importance of sizing the WEB appropriately to ensure mechanical stability of the device within the aneurysm. Without appropriate size selection, compression of the device over time can lead to disappointing occlusion rates. ²⁹

We believe parameters such as height and width result in inaccuracies when describing intracranial aneurysms, which are three-dimensional structures, and thus the development of the ‘+1/-1’ rule is a reflection of the inherent difficulty of precisely sizing a three-dimensional WEB device with two-dimensional measures of aneurysm size. A three-dimensional measure, such as auto-segmented volume, would better approximate aneurysm size and allow greater accuracy in WEB device sizing. Because ellipsoid and cylindrical volumes are derived from linear parameters, these measurements were also found to be less accurate than auto-segmented volumes. Treatment of the first 15 aneurysms in our series was done solely with linear measurements of size and resulted in 5 inappropriately sized WEB devices that were discarded. Volumetric measurements were obtained for the following 14 aneurysms. Among those, only 1 WEB SL device needed to be discarded and replaced with a WEB SLS device, indicating a failure of shape selection rather than size. Using auto-segmented volume measurements, only 34% (10/29) of devices were oversized greater than 20% of the aneurysm volume. Of note, while 21% (6/29) of WEB devices were undersized (less than 20% of the aneurysm volume by auto-segmentation), this may be accounted by the dome tip and lobulations which are not directly filled by the nitinol mesh of the WEB device. Acceptable rates of variability of percent differences of aneurysm and WEB volumes remain to be determined. Nevertheless, oversized or undersized devices in this series adequately covered the aneurysm ostium as determined by DynaCTA and were considered as successful WEB implants.

Because auto-segmented volume is a more precise measure of aneurysm size, as expected, the relationship between aneurysm auto-segmented volume and WEB device volume displayed the highest correlation (r = 0.979), while height × width measurements, although strongly correlated, displayed a lower correlation (r = 0.867) among the parameters evaluated (Table 3). Importantly, the difference between these correlations was statistically significant (p = 0.0007), which suggests that the size of WEB devices is more accurately selected by volume rather than height × width. The difference observed between these two correlations can be accounted by the deformation of the device in three planes as it is deployed within the target aneurysm. Our angiographic follow-up demonstrating 85% complete aneurysm occlusion surpasses other large series, ⁹ in which complete aneurysm occlusion was achieved in only 54% of cases. Although the small sample size and short-term follow-up are significant considerations in our series, these findings further support that volumetric sizing may improve the efficacy of the WEB device. It should be noted that simply using volume measurements to select WEB devices may be inadequate since the discrepancy between WEB volume and aneurysm volume can be very large as described by this analysis. An aneurysm with a height, width, and depth of 6 mm×4 mm×4 mm, respectively would have the same volume as an aneurysm measuring 4 mm×6 mm×4 mm, but may require a different WEB device. Instead, this study supports the use of volumetric measurements to greatly assist in WEB device size selection, rather than as the sole parameter to size selection.

This study has a number of limitations. First, the study is limited by a small number of patients and retrospective design, however this was a preliminary assessment of a novel method of WEB sizing. Second, although the adequacy of immediate aneurysm occlusion was assessed for each implanted WEB device by DynaCTA, it remains to be seen whether our cohort displays similar long-term occlusion rates as patients treated at other institutions. Additionally, the accuracy of volumetric measurements is limited by volume rendering software to appropriately segment a 3D rotational angiogram to obtain volume calculations. In fact, in this cohort we were able to deliver WEB devices which volumes were more than twice the volumes of the initial aneurysm measurements. The intrinsic compressibility of the WEB to axial loading ²⁷ will adjust its volume minimally. The distal and proximal ends of the mesh can partially involute, thereby reducing the volume of the device, which explains how WEBs that were larger than the aneurysm volume were able to fit adequately inside the aneurysmal sac without protrusion into the parent vessel. As a braided mesh, the WEB device has some degree of compressibility such that changes in external pressure (such as pressures applied by the aneurysm walls) can reduce its volume. The compressibility of the device has not been well-characterized in vivo, but many users recommend slight compression of the device to prevent long-term compaction and aneurysm recurrence. Nevertheless, large discrepancies between WEB volumes and aneurysm volumes are likely due to inaccurate volume rendering. Therefore, the three-dimensional volume rendering program available on the Syngo workstation should be validated against other third-party commercial segmentation software for accuracy. In our study, post-implantation volume measurements were unavailable, but may potentially support the intrinsic compressibility of the WEB device which has been observed in our experience. On the other hand, undersized devices may be a result of aneurysm lobulations that increase measured aneurysm volumes but are not directly filled by the device. These irregularities present challenges to WEB sizing, similar to the challenges faced with coil embolization, and require manipulations of the segmentation software and user experience to overcome. Finally, further studies are required to assess the feasibility and efficacy of this method in a larger cohort of patients.

Conclusion

Aneurysm volumes are highly correlated with WEB volumes, with auto-segmentation volumes displaying statistically significant difference against conventional height by width measurements. These results suggest that auto-segmented aneurysm volumes provide an adjuvant method to greatly improve accurate size selection of the WEB device.