Virtual-versus-Real Implantation of Flow Diverters: Clinical Potential and Influence of Vascular Geometry

Discussion

During the past decade, the rising use of intracranial stents and flow-diverter devices for the treatment of intracranial aneurysms has strongly stimulated biomechanical research in the simulation of stent deployment and its impact on the vasculature and hemodynamics. In particular, the device and implantation modeling is constantly refined (including the design and material properties of the device,²¹ its interaction with the vessel,²¹ the deployment procedure,¹⁹ the computational scheme²⁰), increasing the precision of the model at the expense of the computational cost and manual intervention. However, the relevance of each model component is still debated because comparisons with in vivo stent implantations are poorly reported. In the present work, we aimed to fill this gap by comparing virtual and real FDS deployments in a cohort of patients. On the one hand, this work allowed testing the accuracy of the recently proposed deployment model for braided stents.¹⁷ On the other hand, the vessel geometry modifications occurring during the endovascular procedure and their impact on the stent implantation were also investigated.

The virtual deployment was based on some assumptions (ie, constant interwire distance and circular cross-section), which were validated in vitro and in vivo on braided stents.¹⁷ Under these assumptions and knowing PED parameters (interwire distance, wire number, and length and width assessed from manufacturer data or measured experimentally^17,31), the proposed model could predict the main features of real FDSs (length, shape, filament location, porosity, and landing zone) without any free parameters. The measured low stent cross-section eccentricity confirmed the assumed tubular envelope of the virtual stent even if noisy in vivo conebeam CT prevented the characterization of the radiopaque wires of the PED as seen with in vitro measurements.¹⁷

The assessment of the virtual stent radius was shown to play an important role in the prediction accuracy. When based on pre-FDS implantation imaging, this radius was generally underestimated. On the contrary, the radii extracted from post-FDS 3DRAs were all slightly overestimated. These radial differences impacted directly the virtual stent deployment as confirmed with the high correlation between the real/virtual stent radius and length differences. This strong relation showed that the deployment model guarantees an accurate stent prediction when the radius of the stent envelope is correctly assessed as already suggested in Bouillot et al.¹⁷

The differences between real and virtual stents were partly due to the vessel geometry variations of shape and size occurring during the intervention, along with the procedure itself. In addition to the vessel straightening reported by King et al,³² a recurring vasodilation was observed after FDS implantation. These vessel modifications varied strongly from patient to patient and were not confined to the FDS surroundings. This finding indicates that the radial force of the FDS is one among other factors impacting the vessel geometry such as the mechanical impact of the catheter and vasodilator drug (nimodipine) administration. These unpredictable effects were broadly ignored in most of the previous work on virtual deployment and would require additional investigation. Furthermore, various per- and postdelivery mechanical actions performed by the operator can impact the final state of the device. For instance, the improvement of the wall apposition or proximal behavior of the stent requires additional manipulation for expanding the device as shown in Fig 4. These per- and postdelivery manipulations generally result in a compaction and therefore a shortening of the device, which are obviously not taken into account in our model.

Despite the vessel geometry variations and deployment concerns, the virtual deployment based on 3DRA pre-FDS implantation predicted the FDS length with an accuracy of about 10%. Moreover, the virtual FDS length was not biased, contrary to simplified cylindric deployments, thus ensuring accurate predictions when the stent radius is correctly assessed (ie, the vessel geometry does not vary substantially during the intervention). Most interesting, the performance of our virtual deployment was improved when the 8 cases with additional manipulation were removed from the analysis.

The minimal computational cost (only about 1 second on a laptop) and manual intervention (selection of the neck location and stent distal position) of the proposed virtual deployment are appropriate for a potential clinical application to anticipate the device positioning. Furthermore, because the underlying mathematic model is applicable to any braided stent, the virtual deployment can be extended to other types of devices. However, the circular cross-section and constant interwire distance assumptions, which have been experimentally confirmed on PEDs, should be tested with any other devices.

Conceptually, the FDS serves 2 functions in the treatment of aneurysms: It redirects blood flow and it can act as a scaffold for neointimal colonization. The reported issues associated with flow diversion concern nearly exclusively the treatment efficacy and its delayed complications (aneurysm rupture after treatment, side branch occlusion), which are basically related to hemodynamic changes, biologic factors, and medication strategies.^33,34 However very few reports are dedicated to procedural complications and, in particular, the implantation and the deployment concerns of such complex devices. Due to its woven structure and its high wire density, the deployment of the FDS is not a straightforward manipulation. Zanaty et al³⁵ concluded in a recent review that a careful manipulation of the device and a proper deployment reduced the procedure-related complications. Case 16 in Fig 5 illustrates these deployment concerns and their related delayed influence on the parent vessel. An insufficient stent apposition is visible in the inner curvature of the proximal carotid siphon immediately after implantation. Although this insufficient stent apposition is not considered an issue in this kind of procedure, one can see a vessel narrowing in the insufficient stent apposition area on the 6-month follow-up 3D angiogram. Even if the patient remained asymptomatic, the physician would have preferred to avoid such stent proximal behavior during the procedure.

• Download figure
• Open in new tab
• Download powerpoint

Fig 5.

A patient (case 16) with proximal insufficient stent apposition visible on conebeam CT in sagittal (A.1) and coronal (A.2) sections (white arrows). At the 6-month follow-up 3DRA, a slight narrowing is visible in the inner curvature of the internal carotid artery siphon (B.1 and B.2). This narrowing might be a bit overestimated on the 3D rendering (B.2).

Deployment issues related to the landing zone of the proximal end of the stent have already been reported by different in vivo and in vitro experiments conducted by Reymond et al.³⁶ Particularly, they described the accordion effect, which is a protrusion of the stent inside the aneurysm associated with a too-short landing zone. In addition, they observed a transition zone of variable porosity at the neck level associated with an inhomogeneous endothelial tissue colonization of the device scaffold, which could be associated with a treatment failure. Besides, Estrade et al³⁷ reported a case with very bad delayed outcomes related to implantation issues: A too-short landing zone and an oversized device induced a proximal conically deformed stent and a stenosis followed by immediate thromboembolic complications.

On the basis of these observations, we strongly believe that the proposed virtual deployment tool could assist the clinician in avoiding the complications described above. In practice, the deployment prediction together with its possible variability can be assessed preoperatively (ie, as soon as the 3DRA is acquired), providing additional information for optimizing the choice of the device (type, length, and diameter). Furthermore, the prediction of the proximal end of the device is essential for selecting the implantation location, ensuring an adequate landing zone. To improve the prediction accuracy and guide the practitioner perioperatively, this deployment method could be also associated with a live imaging tool. Indeed, 2D-3D registration and live imaging tracking methods could be used to locate the actual device position during the deployment to refine continuously the virtual stent prediction and the associated landing zone. Therefore, the operator could be updated on the final state of the stent while manipulating the device. Finally, our method also provides crucial information such as the filament location and neck porosity for further hemodynamic evaluation with computational fluid dynamics simulations.^{13⇓⇓–16}