Software-based simulation for preprocedural assessment of Tubridge flow diverter sizing: a validation study

Background: The purpose of this single-center retrospective review was to validate the use of a software-based simulation of Tubridge flow diverter (FD) in the treatment of intracranial aneurysms. In a single-center cohort of 17 patients undergoing aneurysm treatment with the Tubridge flow divert, we analyzed their pre- and post-procedural angiographic studies to compare the manufacturer-given nominal length (NL), software calculated simulated length (SL), and the actual measured length (ML) of the flow divert using software. Results: Data for the 3 lengths of all 17 patients treated with Tubridge flow diverts were collected and analyzed in this study. Error discrepancy was calculated by mean squared error (ML to NL 10.64; SL to NL 9.95 p>0.05), mean absolute error (ML to NL 2.64; SL to NL 2.60 p>0.05), and mean error (NL to ML 1.26; SL to ML 2.59 p>0.05). Conclusions: The SL was usually greater than the NL given by the manufacturer, indicating significant change in length in most cases. The residual comparing the ML to the NL was significant, as was when comparing the SL to the NL. The assessment of the Tubridge flow diverter using software simulation is safe and effective and the accurate calculating of FD length contributed to the right-sized FD for optimal placement in intracranial vasculature.

Conclusions: The SL was usually greater than the NL given by the manufacturer, indicating significant change in length in most cases. The residual comparing the ML to the NL was significant, as was when comparing the SL to the NL. The assessment of the Tubridge flow diverter using software simulation is safe and effective and the accurate calculating of FD length contributed to the right-sized FD for optimal placement in intracranial vasculature.

Background
Flow-diversion concept was originally formulated for aneurysm treatment in the 1990s. [1,2] With the development of material technology, several new flow diverter (FD) devices, such as the Pipeline Embolization Device (Covidien, Irvine, California), 4 the Silk flow diverter (Balt Extrusion, Montmorency, France), the Flow-Redirection Endoluminal Device (FRED; MicroVention, Tustin, California), and the Surpass stent (Stryker Neurovascular, Kalamazoo, Michigan) have been available, and animal experiments have been performed to prove their efficacy and safety. [3][4][5] The Tubridge FD, developed by MicroPort NeuroTech, Shanghai, China, is a new type of flow diversion device on the basis of our previous hemodynamic studies of intracranial blood flow, aimed at treating complex aneurysms, such as large and giant aneurysms, and providing more treatment options for neurointerventionists and neurosurgeons. [6] The size of FD is usually determined by the diameter of the distal and proximal parent artery, which might change along the vasculature due to its self-expanding character. [7] There is no accurate method to predict the change in length and exact foreshortening before the deployment of the flow diverter.
Advances computer based modelling tools for flow diverter size or braided selfexpending stent have been reported before, and have not been used before for Tubridge FD sizing. [8,9] We present quantitative data of Tubridge FD sizing using software and the actual FD size after deployment.
Eighteen Tubridge FDs were successfully deployed in all 17 patients (Table 1). A single Tubridge FD was used in 7 cases, Tubridge FD+ coiling in 9 cases (52.9%), and multiple Tubridge FD in one case (5.9%). In the case in which multiple Tubridge FD constructs were used, analysis was limited to the first implanted device.
Periprocedural ischemia was detected in one case (5.9%) in the same territory where the Tubridge FD had been deployed. Six-month imaging follow-up was available for 11/17 patients (64.7%) and revealed residual aneurysm in one patient (9.1%).  The error discrepancy between these measurements was calculated using three standard measures of error ( Table 2

Discussion
Flow diverters were designed to treat complex aneurysms that were difficult to be handled by surgical clipping or conventional endovascular treatments, and have been increasingly used in more than 50 countries. [7] These devices were believed to improve long-term effectiveness, due to the capability to alter intravascular hemodynamics. [7] Preliminary evidence has demonstrated their safety and efficacy for treating complex aneurysms. [6] The Tubridge FD is a stent-like vessel-reconstruction devise designed with a high metal coverage rate and low porosity and was approved by the Ethics Committee of our institution and by the Chinese Food and Drug Administration. This FD is available in various diameters and lengths, which were designed with a pore size of 0.040-0.050mm² at the nominal diameter to provide high metal coverage at the neck of the aneurysm after full opening. The large size Tubridge FD is a braid of 62 nickel-titanium microfilaments and 2 platinum-iridium radio-opaque microfilaments; the small Tubridge is composed of 46 nitinol and 2 platinum-iridium microfilaments. This technique aims to make device selection standardized and to help the physicians in the decision-making during the procedures.
The 3D angiographic image before stent deployment was used to calculate the simulated length using software and the shape of aneurysm and variable diameter around the curves were taken into account. Compared with the nominal length provided by the company, the simulated length was found to be as accurate as the Software-based planning tools that visualize the length of flow diverter have been described. [14][15][16] The first generation of software simulation included the direct placement method, where a uniform stent tube was simply fitted into a patent vessel. [17,18] However, the first generation of software simulation deviates from actual result greatly, because the length of braided stent after implantation is greatly affected by the diameter of blood vessel and poor stent-wall apposition.
Other software-based simulation methods were described, such as finite element method, unstructured embedded grids method and fast virtual stenting method. [15,19] The limitation of previous software solutions has ignored the difference between the intracranial vasculature and vivo models, such as a lack of information regarding wall apposition and the inability to test, in real-time, multiple device dimensions and positions. [19] The assessment and measurement of vascular 11 structure in real time in a real-world setting is the greatest challenge for software simulation, particularly for complex aneurysms.
This is the first clinical study to validate the role of software-assisted simulation in Tubridge FD treatment of intracranial aneurysms. Our results represent an important step toward translation of this novel flow diverter into routine clinical practice and provide effective FD sizing method for the neurointerventionist.

Conclusion
The assessment of the Tubridge FD using software simulation is safe and effective and the accurate calculating of FD length contributed to the right-sized FD for optimal placement in intracranial vasculature.

Methods
Angiographic data of 17 consecutive patients, whom undergoing aneurysm treatment with Tubridge FD between May 2018 and September 2018 in our center, were retrospectively collected. The retrospective study was approved by institutional review board of our hospital, and the informed consents were waived.

Baseline characteristics
Baseline characteristics included patient gender and age, as well as aneurysm location and maximum diameter. We recorded the additional use of embolic material such as coils and whether a single Tubridge or multi-Tubridge construct was used.

Perioperative complications and outcome
Perioperative complications, such as Tubridge deployment failure, occlusion of covered branches, and occurrence of hemorrhage or ischemic stroke, were recorded. Radiologic outcome, including the presence of residual aneurysm perfusion, was recorded for all patients with 6 months imaging follow-up.

Stent length measurement
Three important variables were documented: 1) nominal length of the Tubridge, which was stipulated by the manufacturer; 2) simulated length, which was calculated by the software; and 3) measured length which was measured by a neurointerventionist after deployment. The length of the implanted Tubridge inside the artery was measured from 2D angiographic sequences after operation. The 3D model obtained from the 3DRA images performed before operation was used to generate a centerline. The 3D model was manually registered to the same point of view as the 2D image were acquired. The contrast phase obtained from 2D sequence was used to determine the anatomical location and orientation for a precise alignment between 3D model and 2D image sequence. Finally, the distal and proximal markers of the device were identified on the 2D image and projected on the centerline. The length of the Tubridge FD was calculated as the distance between the distal end and proximal end along the centerline. Biplane acquisitions were available in all cases to choose more clear visibility.

Simulation of FD length
Tubridge FD length was simulated using the braided device foreshortening (BDF) algorithm. A 3D model of the vessel is used to extract the centerline and characterize the local vessel morphology. The computational models used to simulate each device are based on numbers of wires of the stent and the length at two different diameters, which was available from device manufacturer specifications. The obtained model was used to parameterize the simulation following the procedure described by Fernandez [20].
Local morphological descriptors were used to assess length change at each region 13 of the vessel. The distal end position of the simulated FD was matched to the position observed on the 2D angiographic sequences and the proximal one was obtained by running the BDF algorithm. The final length of the FD (simulated length) is the distance between the distal end and the proximal end of the simulated FD along the centerline.

Error assessment between measurements
The simulation was assessed using an error measure. The error between both the measured length and the simulated length was compared to the nominal length stated by the manufacturer using the following criteria: In the above, L is either the measured (M) length or the simulated (S) length, and N is the nominal length stated by the manufacturer.

Statistical Analysis
Differences in baseline characteristics were evaluated with a Student t test, Fisher exact test, and Wilcoxon signed rank test, as applicable. The Statistical Package for the Social Sciences (SPSS, Chicago, USA) was used for statistical analysis, and a pvalue < 0.05 was considered statistically significant. Bland-Altman plot was used to evaluate the consistency of simulated or nominal and measured lengths. These values were charted on a scatterplot, which shows a positive correlation between mea Bland-Altman plot shows the mean difference in length between the measured and simulated