A realistic way to investigate the design, and mechanical properties of flow diverter stents

2.1 Endovascular stent placement and flow evaluation

Aneurysm model

A patient-derived aneurysm model of the left internal carotid artery was produced based on 3D rotational angiographic data (3D RA, Figure 1a) acquired from a patient (female, between 50 - 55 years old). It was obtained as a part of a clinical routine using an X-ray angiographic system (Allura Xper FD20/10 biplane, Version R7.2.9, Philips Healthcare, Netherlands). The data were anonymized before further processing. The protocol to produce the model comprised: 1) segmentation of a vascular lumen from the 3D RA dataset (Figure 1a); 2) optimization of the resulting lumen (removing artifacts: closing holes, separating vessels, (Figure 1b) and 3) model preparation for printing (adding external wall and connectors for integration into the flow loop; 4) printing a digital model with stereolithographic 3D Printer (Clear photoreactive resin, Form 2 Formlabs, USA). A detailed protocol of the model fabrication is submitted for publication. The study was approved by the Ethics Committee at the Faculty of Medicine, Kiel University. Four aneurysm models were printed to place the three FDs and one control. The FDs were placed into the models (Figure 1c) by experienced neuroradiologists.

• Download figure
• Open in new tab

Figure 1:

(a) 2D digital subtraction angiography data (DSA) of a patient (female, between 50 - 55 years old) with left internal carotid artery aneurysm. (b) A vascular lumen was segmented from 3D RA, and the side branches were removed from the original data to simplify the model. (c) Photo of the 3D printed patient-derived model with example deployed stent. The aneurysm sac was cut (red line); this way, the stent surface was visible, and it was subjected to the laser microscope imaging. (d) Illustrative laser microscope image of the stent surface for the porosity and the cell-area calculations. The porosity was calculated as the ratio to open cell-area (sum of yellow areas) to total area (blue), and the cell area of the parallelograms was calculated using the parameters (a, b, ϕ) obtained from the image.

2.2 Porosity and cell-area evaluation

The upper part of the aneurysm sac of the patient-derived models was manually removed (redline, Figure 1c). Thus, making the deployed stent’s surface visible and was subjected to the laser microscope (Model VK-X260K, Keyence, Germany) imaging. The illustrative image shows the deployed 3D stent surface as a 2D projection (Figure 1d). The length of the two adjacent facets (a, b) and angle (ϕ) between them were measured manually (using ImageJ, open-source, Figure 1d) and was calculated using equation 1. The area was calculated for different cells in a single image, and the mean and standard deviation were evaluated. Porosity was assessed as a ratio of open areas (Figure 1d, yellow) to the image’s total area (Figure 1d, blue square), as given in equation 2. After the analysis, the FDs were gently removed from the patient-derived models for further mechanical testing.

2.3 Mechanical testing

2.3.1 Circumferential radial force (RF_c)

RF_C was measured with an in-house-built setup (Figure 2a) during the expansion of the stent. The FDs were 1) crimped to a diameter of 2.2 mm (approx. half of the free diameter), 2) deployed into the apparatus (red dotted circle), and 3) gradually released till free diameter (4– 4.3 mm). The crimping device was made in-house after slightly modifying an available patent by Motsenbocker et al. (US007069794B2) [20]. The device consists of 12 radial blades that are pivoted to move as freely as possible. The ends of the radial blades are connected to a force-ring, which in turn is connected to an assembly of force sensor (100 N, Disynet GmbH, Germany) and linear actuators (M-229.26S & L-306.011100, PI miCos GmbH, Germany). So, when a stent expands, it exerts force on the radial blades-force ring-sensor. Thus, the RF_C was measured as a function of the stent diameter during expansion. Details about the correction factor to accommodate the friction is discussed in the supplementary (S. Figure 1).

• Download figure
• Open in new tab

Figure 2

(a) Scheme of the in-house-built setup to measure the circumferential radial force (RFC). An FD (red dotted circle, red arrow) was crimped to a diameter of 2.2 mm and inserted between radial blades of the apparatus. The FD was gradually expanded in the apparatus. The force exerted on radial blades was recorded as a function of stent diameter. (b) Scheme of the setup to measure the longitudinal radial force (RFL). The front view (a grey area) and a top view of the apparatus are shown. The stent (red dotted line) was placed between two parallel plates; one was fixed and another movable. The stent was compressed longitudinally with a movable plate to a displacement of 2.2 mm from its initial diameter. The force was recorded as a function of the plate displacement. (c) Scheme of the setup to measure the bending force (BF). The stents were deployed in the 3D printed tubes of various diameters (3–4.25 mm; with intervals of 0.25 mm). The tubes were bent to a 14 mm radius of curvature, and the force required to bend was measured. (d) The light microscope was used to generate images of the stent surface for visual cell-angle deployed in 3D printed tubes. Each tube had a hole (red outline) to expose the stent surface, and the angle of the parallelogram (α) of the stent surface was measured.

2.4 Endovascular stents

Three flow diverter stents (FD1-3, details below) with a free diameter of 4.0–4.3 mm and length of approximately 30 mm were considered as trail stents. All dimensions of the devices were chosen by experienced neuroradiologists for the considered aneurysm geometry.

FD1: P64 flow modulation device, Phenox, Germany, 64-NiTi wire braid, visibility through two helical strands along with the entire length, and eight proximal markers [21].

FD2: Derivo embolization device, Acandis, Germany, radiopaque NiTi composite wires with platinum core, and three Platinum-Iridium X-ray markers on both ends [22].

FD3: Silk, Balt extrusion, France, 48 NiTi wires, four radio-opaque markers over the entire length, eight extra Platinum smaller wires in the stent [23].

3.1 Endovascular stent placement and flow evaluation

Flow-sensitive MRI of the 3D printed models of human vasculature with and without stents was successfully acquired (Figure 3). Signal voids likely due to metal artifacts were observed predominantly close to the stents’ surface, such that the MRI signal appeared to be strongly reduced in these areas (Figure 3a, yellow arrows). A qualitatively substantial flow jet entering the aneurysm sac was observed without implants, and in the presence of FDs, the flow jet was eliminated (Figure 3b).

• Download figure
• Open in new tab

Figure 3.

(a) Magnitude, (b) velocity, and (c) pathline reconstructions of the flow through the printed model with and without FDs, acquired with 4D flow MRI (protocol 1). (a) Note the signal voids, yellow arrows, indicating the positions of the stents. (b) The velocity maps of the same slice indicate a significant diversion of the flow. (c) As expected, a quantitative analysis of the data found that the mean velocities were dramatically reduced in the aneurysm sac (ROI A), while it was constant in the parent vessel (ROI B).

The mean velocity was analyzed in the aneurysmal sac and parent vessel (ROI A-B Figure 3c). In the aneurysmal sac, the mean velocity was reduced for all stents after their implantation (s. Table1, without FD: 20.53±13.53 cm/s versus 8.76±8.45 cm/s, 10.35±11.32 cm/s and 8.71±9.02 cm/s for FD1-3 respectively). In the parental vessel, it was almost constant (s. Table1, without FD: 34.89±20.12 cm/s, with FD 1-3: 36.03±17.54 cm/s, 41.27±21.33 cm/s, 36.12±16.74 cm/s, respectively). Note that a large standard deviation is expected, mostly for two reasons: 1. developed flow has a parabolic profile in a circular cross-section with a maximum value at the center and a zero value at the wall; 2. The vessel diameter is not constant across the model, resulting in a change in the velocity to maintain the total flow. Overall, the data indicate a sufficient flow diversion for all stents, suggesting that the stent placement was successful.

3.2 Porosity and cell-area evaluation

The upper part of the aneurysm model was cut-off (yellow arrow) to examine the stent surface (green rectangle, Figure 4a). The area between stent wires (cell-area) and porosity was evaluated at proximal (Pos. 1), central (Pos. 2), and distal (Pos. 3) positions of the stent (blue arrows). The laser microscope image was analyzed to obtain the cell parameters (a, b, and ϕ) (Figure 4b). The stent porosity (Figure 4c) of the FD surface was calculated using equation (2) at three different positions (Pos. 1-3) from these pictures. The porosity of FD1-2 decreases from Pos. 1 to Pos. 3 by approximately 5.5 %, whereas for FD3, it increases by 4 %. The cell-area of the FD surface were calculated using equation (1) at the same positions (Pos. 1-3) (Figure 4d). A slight decrease in the cell-area was observed for FD1-3. For better quantitative comparison, the cell-area at the Pos. 2-3 was normalized by the area at Pos. 1 (Figure 4e). The highest decrease in the area from Pos. 1-3 was yielded by FD1 (24 %), followed by FD2 (13 %) and then FD3 (5 %). The error bars indicate the standard deviation of the different cell-areas calculated in the same image.

• Download figure
• Open in new tab

Figure 4

(a) Photo of the 3D printed patient-derived model with a deployed stent. The aneurysm sac was cut off (yellow arrow) to provide access for a laser microscope. The images were taken at three different positions (Pos. 1-3), as indicated by blue arrows. (b) The laser microscope image of the stent surface was used to calculate porosity and the cell-area using the parameters (a, b, Ø). (c) Resulting porosity at positions 1 -3. Interestingly, the porosity was reduced for FD1-2 from Pos. 1 to Pos. 3. In contrast, for FD3, the porosity increased for the same. (d) Measured cell-area at three different positions (Pos. 1 -3). Note that the cell area was reduced slightly from proximal to distal positions. (e) The calculated cell-area normalized to value at Pos. 1, however, showed that the decrease from Pos.1 to Pos.3. The error bars indicate the standard deviation for cell-areas calculated on the image obtained at a specific position.

3.3 Mechanical testing

Circumferential radial force (RF_C), Longitudinal radial force (RF_L), and Bending force (BF) analysis

The RF_C was measured as FDs expanded to their free diameters, which was slightly different for each FD (4.0–4.3 mm). The RF_C exerted by all the FDs during expansion was very similar (Figure 5a). Moreover, the FDs showed two varying regimes of expansion, which are marked in yellow. The lower diameter regime has a negative slope of 2.38 N/mm, an order of magnitude higher than the slope at a higher diameter regime (0.20 N/mm).

• Download figure
• Open in new tab

Figure 5

(a) Circumferential radial force (RFC) exhibited two different linear regimes during the expansion (indicated in yellow), with FD2 exerting the most and FD3 exerting the least force. (b) Longitudinal radial force (RFL) was increasing approximately linearly during the compression. A rapid increase feature (red arrow) is consistent with all the three repetitions, suggesting that it might be caused by the particular design of the FD2 stent. (c) Bending force (BF) required to bend FD1 – FD3 to a bending radius of 14 mm when deployed in tubes with diameters 3 – 4.25 mm (three-point-bending test). (d) Normalized BF for FD1 and FD2 with the value of FD3 at every diameter is shown. Note that the maximum BF required for all the FDs is 3.75 mm diameter and gradually decreases for the lower and higher diameters. From the normalized graph, it is evident that the force required to bend FD2 was more than twice compared to FD3, and for FD1, it was approximately 1.5 times higherthan FD3. Each data point was measured three times, and the error bars indicate the standard deviation. (e) Visual cell-angle analysis

Figure 2) of the unit cells on the FD surface deployed in different diameters. Each diameter was imaged 3 three times, many cells were measured from each image, and the standard deviation is as indicated.

The RF_L was plotted as a function of plate displacement (Figure 5b). The force exerted by FD2 (1.58 N) is higher by a factor of two compared to FD1 (0.96 N) and FD3 (0.77 N) at a plate displacement of 2 mm. A rapid increase of the RF_L at 1.2 mm of FD2 compression (red arrow) was reproducible over three measurements, suggesting that this could be arising due to the particular design of FD2.

The BF was measured as a force required to bend the flexible tubes with deployed FD1-3 (Figure 5c) to a radius of 14 mm. Note that all the FDs showed a maximum BF at a 3.75 mm diameter and gradually decreasing for the lower and higher diameters. The normalized BF of FD1 and FD2 with the value of FD3 at every diameter (Figure 5d) was evaluated. The force required to bend FD2 was more than double compared to FD3, and for FD1, it was approximately 1.5 times higher than FD3. Additionally, the cell-angle (α) was evaluated between the struts (Figure 3f) for FD 1-3, depending on the tube diameter (3–4.25 mm, Figure 5e). We noticed a general trend that the angle (α) increases as the diameter increases for all the FDs.

4.1 Endovascular stent placement and flow evaluation

4D flow MRI allowed to access the flow diversion in the patient-derived models (Figure 3a-c). A 50-58 % reduction in mean intra-aneurysmal velocity was detected, which shows that the devices are successfully placed along the aneurysm’s neck.

However, 4D flow MRI is prone to metal-induced artifacts. Earlier, the influence of metal artifacts in the presence of the braided stents (FD1 and FD3) on flow estimation was investigated [24]. The stents were placed in the straight tubes; first, it was shown that the static background was not affected by the stents, and, second, the mean velocity magnitude was almost the same along the tube in case of parallel orientation of the stent to the magnetic field. In our study, the model was oriented along the direction of the magnetic field; thus, the stent was mostly aligned parallel to the magnetic field. Another technique, e.g., single-point 4D flow MRI [25], can be more robust toward the artifacts, but the scan times are too long for in vivo applications.

Other shortcomings of 4D flow MRI often include insufficient spatial resolution for small vessels and aneurysms, limited sensitivity to the low velocities, and non-repetitive flow patterns. Higher fields [26] and acceleration techniques [27] can overcome some of these issues; real-time MRI may be used to image non-repetitive flow [28]. The flow diversion can be investigated with higher accuracy by blood simulation or particle-image velocimetry. However, the latter works exclusively in vitro and has no equivalent in vivo. In contrast, 4D flow MRI examination can be the same for both in vitro and in vivo.

4.2 Porosity and cell-area evaluation

The porosity (Figure 4a) was almost constant (70-80 %) over Pos. 1-3 for all the FDs. The variation can be due to the change in the vessel’s diameter from Pos. 1 (3.67 mm) to Pos. 3 (4.25 mm), diameters obtained from 3D model. The stent cell-area decreases from Pos.1-3 such a variation can be caused due to two reasons—first, the variation in the model vessel diameter. Second, the compression/foreshortening of the stent during placement, which can be done by the physician to achieve a denser mesh near the neck of the aneurysm.

The proposed method allowed to evaluate the porosity and cell-area in complex models. To our knowledge, they were evaluated previously using simplified models of vessels (straight tubes [10,16]). It is essential to use a laser microscope instead of a light microscope as the latter can only provide good focus only to a specific focal point. When placed in curved geometries, the stent surface can have structures at multiple focal points, making it challenging to use a light microscope. On the other hand, the laser microscope scans for the surface from an absolute minimum to maximum interval in z-direction and generates a 2D stent surface image for analysis.

This technique’s main shortcoming is that the laser scan images differ slightly compared to the actual data. The error arises as the placed stent is in 3D, and the acquired laser image is in 2D; this 2D projection varies slightly with the degree of the stent curvature. Such error scale linearly with the curvature of the stent surface. The higher the degree of bending, the larger is the error. The error was minimized in the current study by taking the scans from a small area, limiting the degree of bending. More analysis and studies are to be done to better account for these errors. The histological analysis of animal models is another way to evaluate porosity in complicated geometries [29]. Though these models give additional information about the cell-growth on the stents, their main limitation is that one animal model is required to test the placement of one stent by one interventionalist.

4.3 Mechanical testing

The previous studies [10,11] reported on braided FDs for RF_C measurements were done at a particular diameter (85 % of free diameter). In the current analysis, the RF_C was measured first, as a function of diameter (2.2 mm to free diameter) and second, at 85 % of free diameter (Table 1). The evaluation as a function of diameter could give a better estimate of the force if the FDs were deployed in a changing diameter scenario (distal to proximal), often the case in the human brain [13]. In comparison with literature (Table 1), the resulting RF_C is slightly less than it was reported in the previous studies with braided FDs; this could be attributed to the differences in the stent properties (e.g., geometrical, design, and material). The reliability and reproducibility of this method are trustworthy, as the error bars indicate minimal spread. Commercially RFc measuring devices (also called a crimping machine) are relatively well established. Providing the data of different flow diverters for a range of diameters instead of at a particular value can help compare different devices.

In general, brain vessels have a noncircular cross-section; thus, it is essential to evaluate not only RF_C but also RF_L. The effect of RF_L investigations for stent ovalization shows a similar trend between different stents. The analysis shows that the RF_L varies continuously with the plate compression. Compared to literature (Table 1) of braided FDs, our data has slightly higher values. The same argument can be considered as before, as the previous studies were focused on different devices. The reliability and reproducibility of these tests are good with a mean error of 0.032 N (calculated as a mean of all error bars) and with a maximum error of 0.081 N.

Moreover, we observed that the forces reduce and increase for diameters higher and lower than 3.75 mm for BF analysis. The normalized BF graph shows the force exerted between the different stents FD 1-3. The visual cell-angle at 3.75 mm diameter shows that high force is observed when α is approximately 90 degrees. The rectangular cell geometry (when α = 90 deg) could be one reason for this high value. However, this was seen only in FD1 and FD3, and the FD2 showed a significantly higher angle at this diameter. This has to be further verified with reliable computer simulations and more extensive test specimen sizes. The BF measurements prove to be reliable and repeatable for different stents. Both the RF_L and BF were reported earlier at specific diameters, and they can be further extended easily to measure at different diameters. We show that one can slightly modify a commonly available characterization setup like tensile/universal testing machine to obtain the data at varying diameters with good reliability.

To summarize, we applied and discussed different methods to characterize braided stents. This data can serve as guidelines for future design development when comparing to state-of-the-art. The bigger question unanswered is if the porosity, cell-area, and mechanical parameters play any role in the patient treatment outcomes. Generating such data sets and comparing it with long-term clinical performance could give us a better chance of understanding this question. To fulfill this, the data sets need to be as close to the real-life scenario as possible. In the current study, evaluating the porosity and cell-area in patient-derived models and measuring the mechanical forces in a range of diameters brings it a one-step closer to the real-case. Therefore, these methods serve as guidelines for further developments in the testing of braided FDs.