Int J CARS DOI 10.1007/s11548-013-0943-2

ORIGINAL ARTICLE

Stent graft visualization and planning tool for endovascular surgery using finite element analysis S. von Sachsen · B. Senf · O. Burgert · J. Meixensberger · H. J. Florek · F. W. Mohr · C. D. Etz

Received: 29 April 2013 / Accepted: 3 September 2013 © CARS 2013

Abstract Purpose A new approach to optimize stent graft selection for endovascular aortic repair is the use of finite element analysis. Once the finite element model is created and solved, a software module is needed to view the simulation results in the clinical work environment. A new tool for interpretation of simulation results, named Medical Postprocessor, that enables comparison of different stent graft configurations and products was designed, implemented and tested. Methods Aortic endovascular stent graft ring forces and sealing states in the vessel landing zone of three different configurations were provided in a surgical planning software using the Medical Imaging Interaction Tool Kit (MITK) software system. For data interpretation, software modules for 2D and 3D presentations were implemented. Ten surgeons evaluated the software features of the Medical Postprocessor. These

surgeons performed usability tests and answered questionnaires based on their experience with the system. Results The Medical Postprocessor visualization system enabled vascular surgeons to determine the configuration with the highest overall fixation force in 16 ± 6 s, best proximal sealing in 56 ± 24 s and highest proximal fixation force in 38 ± 12 s. The majority considered the multiformat data provided helpful and found the Medical Postprocessor to be an efficient decision support system for stent graft selection. The evaluation of the user interface results in an ISONORMconform user interface (113.5 points). Conclusion The Medical Postprocessor visualization software tool for analyzing stent graft properties was evaluated by vascular surgeons. The results show that the software can assist the interpretation of simulation results to optimize stent graft configuration and sizing.

Electronic supplementary material The online version of this article (doi:10.1007/s11548-013-0943-2) contains supplementary material, which is available to authorized users.

Keywords Medical Postprocessor · Simulation-based treatment planning · Biomedical visualization · Implant planning · Finite element analysis · Stent graft · EVAR

S. von Sachsen (B) · J. Meixensberger · F. W. Mohr Innovation Center Computer Assisted Surgery (ICCAS), University of Leipzig, Leipzig, Germany e-mail: [email protected] B. Senf Fraunhofer Institute for Machine Tools and Forming Technology IWU, Dresden, Germany O. Burgert Department of Informatic, Reutlingen University, Reutlingen, Germany H. J. Florek Weisseritztalkliniken GmbH Freital, Freital, Germany C. D. Etz Mount Sinai Hospital, New York, NY, USA

Introduction The first stent graft for endovascular exclusion of aortic aneurysm was implanted in 1986 by Volodos (Ukraine) [1]. In Argentina and the USA, the new technology was used in 1991 by Parodi [2] and Dake [3] for the first time. In Germany, introduction of stent graft implantation took place 1994 by Stelter [4]. According to the German Society of Vascular Surgery’s report, Abdominal Aortic Aneurysm, there is an increasing use of endovascular aortic reconstructions (EVAR) in Germany [5]. While in 1999–2002, only 17 % of all operations for exclusion of abdominal aortic aneurysm were performed endovascularly, the percentage rose to 45 %

123

Int J CARS Fig. 1 Left volume rendering of abdominal aortic aneurysm with calcified plaque (Screenshot Software Osirix), right AnacondaTM stent graft system with bifurcation body (center) and legs [6]

of all interventions in 2008 (2,091 stent graft implantations in certified vascular centers). The majority of EVAR treatments are those of abdominal aortic aneurysms (Fig. 1, left side) for which it is not unusual to use modular systems with three or more components (Fig. 1, right side). In German hospitals, there are currently approximately seven products used, which differ significantly from each other in design, material and proximal anchorage mechanism. Postoperative complications, such as stent graft migration (sliding of the graft >10 mm) and endoleak type I (insufficient sealing in the landing zones), have been decreased by material and design optimization of stent grafts. However, the aforementioned complications can still be observed (migration 5 %, endoleak 7 % [7]). It can be assumed that the occurrence of migration and endoleak type I is associated with the selection of stent graft design and determination of oversizing, which depends mainly on the experience of the vascular surgeon. Currently available planning tools like Osirix [8] and TeraRecon [9] are limited to a measurement function and 3D visualization of the vessel. There exists no commercial planning system that enables a quantitative evaluation of stent grafts and a comparison of different devices and configurations. The use of finite element method (FEM) is considered a helpful approach for simulation of stent graft– vessel interaction. Through this method, stent graft forces and stent graft fitting in patient-specific vessel can be numerically approximated. With the aim to assist the vascular surgeon with indication of treatment, Raghavan et al. presented a 3D finite element (FE) model of an abdominal aortic aneurysm for better estimation of rupture risk [10]. Meanwhile, the company VASCOPS GmbH (Austria) provides a commercial software for analyzing rupture risk using calculated wall stresses and a so-called rupture potential index (RPI) [11]. Due to the fast development of modern hardware and software, it is now possible to use patient-specific calculation models, e.g., based on computed tomography angiography. This makes the use of structural and fluid mechanical simulations more and more

123

interesting for answering medical questions. In the field of EVAR, numerical analysis of the flow characteristics before and after implantation of abdominal stent grafts [12,13], and determination of stent graft displacement forces [12,14] were presented. A new movement can be observed toward development of simulation models for use in daily clinical routine for patient-specific optimization of stent graft selection [15,16]. It is important to integrate the surgeon in the model development process for determination of important model parameters and suitable software functionalities to allow result interpretation. For this purpose, a software module is needed that provides vascular surgeons with access to simulation results. Therefore, the presented work introduces helpful evaluation values, appropriate data presentation forms and user friendly interaction methods.

Materials and methods After performing finite element analysis (FEA) of a stent graft candidate, the results have to be interpreted using a visualization software tool such as the Medical Postprocessor. This new software module was integrated into the FE analysis processing chain to visualize the patient-specific results, illustrated in Fig. 2. Finite element analysis Determination of helpful FE result values For determination of clinical requirements on a FE model and definition of helpful result values for evaluation of stent graft properties, participative methods like questionnaire and cooperative prototyping was used. This realizes the integration of a vascular surgeon in the development process of the simulation-based planning approach. Questionnaire responses show that over 90 % of vascular surgeons interviewed consider stent graft fixation force and contact state

Int J CARS

Fig. 2 Process chain for integration of FE results in Medical Postprocessor

between vessel wall, plaque and stent rings as helpful values for improving stent graft oversizing and positioning [17].

radial force of a stent graft ring (FR )rad , the following formula is suggested:

Fixation force

(FR )rad = f (Stent O , Blood p , V esselm , Plaquem , Stentm )

(2) According to Gebert de Uhlenbrock, stent fixation force FStent is calculated from radial stent force stentForce, blood pressure pressureForce and a friction coefficient, µ (0.4), as follows [18]: FStent = µ ∗ (stent For ce + pr essur eFor ce)radial

(1)

Formula (1) of Gebert de Uhlenbrock was defined in experimental studies. For numerical simulation of stent graft forces a parameter study was performed by Senf et al. to identify influencing parameters for calculation of stent radial force [19]. Based on this, formula FR was defined which considers oversized stent ring (Stent o ), and material properties of plaque Plaquem , vessel wall V esselm , and ring Stent m , as well as blood pressure. For numerical approximation of the

with the friction coefficient µ of Gebert de Uhlenbrock and the afore-introduced radial ring force (FR )rad , the fixation force of a stent graft ring FR can be calculated as follows: FR = (FR )rad ∗ µ

(3)

For evaluation of fixation potential in landing zones, the evaluation values fixation force stent graft proximal, FSGpr ox , fixation force stent graft distal left, FSGdist L , and fixation force stent graft distal right, FSGdist R , are introduced (see Fig. 3), which can be calculated by adding all ring forces of respective landing zones, proximal FRpr ox , distal right FRdist R and distal left FRdist L :

123

Int J CARS integration points

stent graft ring vessel wall calcified plaque

CE = 1 contact exists

CE = 0 no contact

Fig. 4 Calculation of contact states between stent graft rings, plaque and vessel wall

Kleinstreuer et al. found that graft material has no significant influence on mechanical response of stent wires [23]. Consequently, the covering of the stent graft is not considered within this study. State of FE contact elements To provide a better estimation of endoleak type I risk, the calculation of contact states for all elements of stent graft rings is suggested. Therefore, stent graft rings have to be meshed with contact elements, and vessel wall and plaque have to be meshed with target elements. The state of each contact element is binarized and can have a contact state with value 0 (no contact), or with value 1 (contact exists) (see Fig. 4). Data acquisition

Fig. 3 Fixation and flow forces in stent graft landing zones

FSGpr ox = FSGdist R = FSGdist L =

n  n=1 n  n=1 n 

(FRpr ox )i

(4)

(FRdist R )i

(5)

(FRdist L )i

(6)

n=1

The flow forces FB F pr ox (blood flow force proximal),  FB Fdist L (blood flow force distal left) and FB Fdist R (blood flow force distal right), which are necessary for determination of migration risk, are to be calculated using a hemodynamic model for respective landing zones [20]. The focus of the presented work is on the provision of methods for interpretation of calculated stent graft properties generated by simulation for evaluation of stent grafts. Therefore, it is sufficient to generate exemplary result values using a structural mechanical FE model and the use of proximal and distal flow forces from literature [21,22].

123

The process chain to evaluate stent graft properties (see Fig. 2) starts with data acquisition and associated segmentation of patient-specific model parts. Vessel geometry: To diagnose an abdominal aortic aneurysm, a CTA scan is performed as part of routine clinical procedures. The CTA data used for generation of vessel and calcified plaque geometry have 643 slices with a resolution of (512 × 512) px and a voxel size of (0.76 × 0.76 × 0.75) mm. As it is not possible to determine vessel wall thickness in CTA slices, data from experiments have to be used. The parameters describing the material properties of relevant parts determined by Gao [24] (vessel) and Holzapfel [25] (calcified plaque) were used for the same reason. Stent graft: For modeling stent graft components (bifurcation body and legs), engineering data were provided by the manufacturer. Description of stent rings’ Nitinol material was enabled by a stress–strain diagram provided by the manufacturer. Data preprocessing Based on CTA data, surface models of vessel parts were generated using the segmentation software Mimics [26]. The vessel wall and plaque have to share polygons for which a surface

Int J CARS

model is generated in 3-matic that exists of a non-manifold mesh [27]. Furthermore, 3-matic is used for mesh optimization and for generation of a volume mesh which exists of tetrahedrons. An optimal, tetrahedron-based FE mesh has polygons with a dihedral angle less than 165 degrees and an aspect ratio of edge length of 1:20 (FE software ANSYS, manual [28]). FE simulation The volume model of the vessel generated in 3-matic was exported as *.cdb-file (ANSYS-scene) and thus can be loaded directly into the FE-software ANSYS. The superelastic Nitinol stent rings of the Anaconda stent graft [6] were modeled as shape memory alloy objects existing of hexahedrons and placed in the landing zone using ANSYS features. After stent modeling, contact elements on the inner vessel and outer stent surface are determined. The simulation of the interaction between stent graft rings, plaque and vessel wall using a patient-specific FE model is a self-contained, complex research field and was not the focus of the presented work. In fact, the FEA performed for the presented work serves primarily for the generation of exemplary result values. Therefore, it is sufficient to generate radial forces and deformation states performing a compression of the stent ring objects in the vessel model. For the demonstration of tight and leaky regions in the landing zones, a subset of contact elements of the rings were assigned the contact state 1 (good sealing). The result values were generated for three different stent graft configurations. Export EVAR relevant result data The export of relevant geometry and result values (radial force, contact state) from FE software was performed using ANSYS-specific scripting language. The vessel wall thickness is not relevant for result interpretation and the visualization of the inner and outer vessel wall geometry would also make it difficult to see the plaque and stent rings inside the vessel when using transparent objects. Therefore, the outer vessel wall geometry was not exported. Another optimizing step was the export of only the corner points of the tetrahedron elements used in the FE software. Thus, the result files contain 4-node tetrahedrons for the vessel and plaque instead of 10-node tetrahedrons used in the FE software for a more precise simulation. To provide the planning relevant result values on a 3D model, the following data were exported: 3D geometry of vessel wall, plaque and stent ring: • • • •

Element type Number of nodes and elements Node numbers with appropriate coordinates Element numbers with appropriate node numbers.

Results on stent rings: • Element numbers of stent ring mesh with appropriate radial forces • Number of contact elements of stent ring mesh • Element numbers of stent ring mesh with contact state 1. To provide result data separately for proximal and distal landing zones, the result file starts with all proximal ring values and then continues with all distal ring values. The values for distal rings were also divided in results for left and right stent graft rings. Software framework: Medical Postprocessor Data processing FE postprocessing (analysis of simulation results) in a medical planning software should enable an explicit visualization of data steered by user interactions. Therefore, an import routine was implemented that generates data container to enable a separate storage of geometry and result values. For data structuring and storage, the Visualization Toolkit (VTK) was used [29]. The planning-relevant FE result consists of scalar values, which can be presented as color-coded mesh nodes using a corresponding color table. Two mapper modules were implemented, which visualize stent forces and contact states based on user input. For presentation of force distribution on the stent ring, a color transfer function is used with a color table of the FE simulation software ANSYS, which contains eight color items c1, . . . , c8 ∈ N. Mesh nodes were colored by mapping force value F ∈ Q according to color table RG B1, which is defined as: RG B1 = [c1, c2, c3, c4, c5, c6, c7, c8]

(7)

Particular color range CR can be calculated using the value range of simulation results and amount of colors in the color table: CR =

rangeRadialForces amountOfColors

(8)

Based on calculated color range, the color transfer function values TFVi were defined TFVi = minValue + (i − 1) ∗ CR

(9)

and assigned to the particular color object ci , of color table RG B1 by initialization of the color transfer function CTF: for (int i = 1; i < 8; i ++) { }

CTF → Add RG B Point (TFVi , ci ) CTF → Add RG B Point (TFV(i+1) , ci )

123

Int J CARS Fig. 5 Medical Postprocessor navigation modules

123

Int J CARS

Contact state C O N T ST AT ∈ {0, 1} of stent graft ring elements was visualized per mesh node according to defined color table RG B2. They are represented in blue and white for sealed and leaky state, respectively:   blue if C O N T ST AT = 1] RG B2 = (10) white if C O N T ST AT = 0 Data analysis The focus of the Medical Postprocessor is on a quantitative and qualitative analysis of stent graft properties and a comparison of multiple stent graft devices and configurations. Quantitative analysis of simulation results refers to overall fixation forces, proximal and distal fixation forces, ring forces and estimation of migration risk. Therefore, explicit indication of calculated forces is given and results are presented using bar charts or overview graphics. Qualitative analysis refers to sealing state and stent graft location in the vessel. The analysis for this purpose is performed using the 3D-Model and is dependent on visual impression of the surgeon. User interface Based on determined user requirements [17], a graphical user interface for analysis of FE result values was developed. The Medical Imaging Interaction Tool Kit (MITK) serves as a development platform [30]. A plugin was implemented which perspectively contains functions for accompanying the whole intervention planning process which consists of the following six work steps: 1. Reading patient data, 2. Measuring vessel diameters and lengths, 3. Stent selection/configuration, 4. Virtual implantation, 5. Simulation and 6. Result interpretation. The focus of the presented work is on result interpretation, for which the Medical Postprocessor was developed. It consists of a navigation area and 2D data representation window on the left side and a window for interaction with the 3D vessel model on the right side (see Fig. 5). A navigation map is available which provides a zooming feature for all landing zones.

The navigation menu of the analysis module is divided into four categories: • • • •

System’s configuration proposal Evaluation of single configuration Comparison of multiple configurations Analysis of migration risk.

The user can toggle between all evaluation modules. An overview of possible workflows when using the mentioned modules is shown as an activity diagram in Fig. 6. Configuration proposal Based on three generated configurations, which differ in proximal/distal diameter and leg length, a configuration proposal is made by the system. Selection criterion was the highest overall fixation force. Several works have studied drag forces of stent grafts using in vitro experiments on animal arteries or silicon-made vessels [31,32]. Therein, drag forces of stent grafts were compared, which suggests that drag forces are currently used for quantitative evaluation of stent graft devices. The analysis module provides a comparison of results for the proposed configuration with other configurations (see Fig. 7). The data in Fig. 7 show that the fixation force of configuration 2 (config2) is 3 % higher than configuration 1 (config1), and 13 % higher than configuration 3 (config3). This configuration proposal is exclusively based on the overall fixation forces, and sealing state is not considered. Single configuration evaluation The module for evaluation of a single configuration enables selection of a configuration using a combo box element. With this module, proximal and distal force values are displayed. In the 3D view, proximal rings are visualized in green and distal rings in yellow. Activation of the check box per ring shows force values of each ring in a bar chart. This data enable an evaluation of the ring’s contribution to stent graft fixation. Figure 8 shows the user interface with module Single Configuration Evaluation activated.

Analysis-ToolBox Configuration comparison The Analysis-ToolBox contains basic elements for selective result visualization. Thus, the user has the possibility to explicitly visualize forces or sealing state on the 3D object. Visualization of forces and contact states can be steered within the first three modules. The color legend helps by classifying visualized forces on the 3D model in a color range (blue = lowest value, red = highest value). Furthermore, a legend is provided for visualizing the contact state. According to this, all rings with good sealing are shown in blue color.

The module for stent graft comparison shows all generated configurations with their associated components (see Fig. 9). All calculated overall fixation forces are listed in a table. The user can steer parallel visualization in the 3D view using check boxes and thus perform further configuration comparison (Fig. 9). The user can interact with each of the three models in the respective window using mouse functions (left mouse button: rotate, mouse wheel: zoom).

123

Int J CARS

Fig. 6 Activity diagram for Medical Postprocessor

Evaluation of sealing potential Evaluation of endoleak type I risk has to be performed visually by the surgeon (see Fig. 9). Blue-colored areas of stent graft rings indicate good sealing, whereas white areas on the outer surface of the stent graft mark endoleak type I areas. Migration risk analysis An automated evaluation of calculated fixation forces in the different landing zones is possible by comparing them with numerically determined flow forces. Based on this, it is possible to assign stent grafts to classified risk groups (see Table 1)

123

and show their potential using colored boxes in landing zones (see Fig. 10). Evaluation of Medical Postprocessor Usability test Ten vascular surgeons from 10 hospitals across Germany participated in the Medical Postprocessor evaluation. The software was demonstrated using a notebook with Intel Core i7-720 QM processor, 1.6 GHz, 6MB L3 cache, ATI Mobility Radeon, 15.6”, 16:9 HD LCD and external mouse. Using the module for evaluation of FE results, surgeons have to deter-

Int J CARS

Fig. 7 Configuration proposal: left configuration components and comparison of fixation force, right visualization of radial force distribution on distal stent graft rings

Fig. 8 Single configuration evaluation

123

Int J CARS

Fig. 9 Configuration comparison: data table and parallel visualization of contact states for three configurations (blue = sealed)

Questionnaire II: Medical Postprocessor usability

Table 1 Classification of migration risk Risk class “Migration”

Mathematical precondition

Low risk

Fixation force > flow force

High risk

Fixation force < flow force

mine the one with best proximal sealing potential, highest proximal fixation force and highest overall fixation force, for three exemplarily generated data sets. Time for decision making was recorded for 8 of 10 sessions, which were uninterrupted.

Questionnaire I: Medical Postprocessor features On the basis of a questionnaire with 26 items, additional information regarding planning habits of surgeons was captured (the questionnaire can be found in the online appendix). Furthermore, surgeons evaluated the modules concerning planning assistance, data structure, data visualization and data interpretability. Statistical analysis of questionnaire data was performed using the software SPSS Statistics (SPSS Inc.).

123

During the development of user interfaces, the international ISO standard DIN EN ISO 9241-110 Dialogue principals has to be considered to evaluate usability [33]. Therefore, the questionnaire of Prümper was employed [34]. This questionnaire enables a concrete evaluation of usability via the criteria suitability of the task, suitability of learning, individualization, conformity, self-descriptiveness, controlability and error tolerance. The ISO standard questionnaire of Prümper et al. is considered to be a useful and formative evaluation method and can be used in context with a participative design concept to determine concrete software flaws [35]. The questionnaire contains 21 items which are answered using a bipolar format. Possible answers range between very bad and very good and are gradated in seven steps. For questionnaire evaluation, the matrix provided by Prümper et al. was used [36]. Results Usability test After an introduction to the analysis module, the surgeons work on the defined task. Results of time measurements for

Int J CARS

Fig. 10 Migration risk analysis

n = 8 surgeons are presented in Fig. 11. The boxes show the middle 50 % of the measured times needed. The line inside the boxes is the median. The ends of the vertical lines indicate the minimum and maximum values. Highest average time needed was measured for determination of the configuration with best proximal sealing (56±24 s). Lowest average time needed was for determination of the configuration with highest overall fixation force (16 ± 6 s). The average time for identification of the configuration with highest proximal fixation force was 38 ± 12 s. Questionnaire I: Medical Postprocessor features Questionnaires for evaluating software features, presentation forms and usability (ISO Standard questionnaire) were answered by ten vascular surgeons from ten different hospitals (100 % male, 50 % chief physician, 50 % senior physician). The respondents were aged between 33 and 63 years (M (mean) = 48 years, SD = 9 years). The experience in the field of EVAR ranged between 4 and 18 years (M = 11 years, SD = 6 years). The amount of stent graft interven-

Fig. 11 Time measurement for configuration determination (n = 8)

tions performed by each surgeon varied between 30 and 1000 (M = 252 interventions, SD = 286 interventions). The majority considered the methods provided for 2D and 3D data presentation helpful. Percental comparison of system’s configuration proposal with other configurations was rated as advantageous by 80 % of the surgeons (see Fig. 12).

123

Int J CARS Fig. 12 Evaluation of 2D data presentation (n = 10)

Fig. 13 Evaluation of 3D model presentation

Comparison of fixation and flow forces for better estimation of migration risk was considered appropriate by 90 % (see Fig. 12). Based on this, a color-coded overview of landing zones for presenting migration risk values for each configuration was developed, which was considered helpful by 100 % of the respondents (see Fig. 12). In this context, the use of well-known colors from the traffic system are assessed appropriate, since they are easily comprehensible for a classification of good or bad configurations. Visualization of fixation forces was evaluated as satisfying by 90 % of the surgeons (see Fig. 13). The binary-coded visualization for identifying potential endoleak type I areas in the 3D model was greatly appreciated (see Fig. 13), because it shows the surgeon which stent graft has highest risk for an appearance of endoleak type I. Separate indication/visualization of proximal and distal forces is considered helpful by 100 % of the surgeons (see Fig. 13). A parallel presentation of multiple configurations in 3D view is an added value from the stance of 70 %, for example, for a better comparison of sealing potential (see Fig. 13). All participants

123

Fig. 14 Evaluation Medical Postprocessor features

rated visualization of thrombus to be as important as visualization of plaque, and therefore as an important point for future FE modeling (see Fig. 13). 80 % of the test persons consider available features helpful for the evaluation of a single stent graft configuration (see Fig. 14); 70 % of the respondents consider it possible

Int J CARS

Fig. 15 Use potential of Medical Postprocessor

to perform a device comparison and 90 % find it possible to determine fixing rings (see Fig. 14). In summary, the evaluators see the field of use for the Medical Postprocessor in an improved estimation of migration and endoleak type I risk, in an efficient decision support referring stent graft selection and in an improved estimation of long-term course as well (see Fig. 15). The questionnaire which was used accompanying the usability test contains also general questions referring the planning habits. Together with the evaluation of the first questionnaire from April 2011 [17] for definition of software requirements, it was possible to collect information about the stent graft planning time from n =18 vascular surgeons with different work experience. The average time needed is 37.7 ± 25.4 min. It should be noted that the time includes all work steps from analyzing the CTA data and measurement to selection of stent graft components from the product catalogue. Currently there exists no method to figure out the effects of different stent graft configurations, and therefore, selection is based on measurement values, vessel morphology and experiences of the vascular surgeon. Questionnaire II: Medical Postprocessor usability Evaluation of the ISO standard inquiry using the evaluation matrix of Prümper et al. results in an ISO standard value, which enables the grading of software usability. There exist four valuation levels: 21–50 points: modifications are absolutely necessary, 51–82: need for action, 83–114: Everything is all right, 115–147: software is perfectly adjusted for the user. Evaluation of n = 10 questionnaires results in an ISO standard value of 113.5 points for the Medical Postprocessor and in accordance with the evaluation matrix of Prümper et al., it can be classified in the second highest stage. With this ISO standard value, there exists

a good usability and there is no reason for user interface modification. The lowest value is determined for the principle individuali zation (¯x = 4.6) and the highest value for the principle suitability learning (¯x = 5.9) (see table 2 of online appendix). Two-thirds of principles found straightforward acceptance by >50 % of the surgeons (agree or fully agree). An overview of the results for all 21 usability principles can be found in the online appendix. Discussion Usability test and questionnaire results show that defined evaluation values for comparing stent graft configurations/ products are helpful, and 2D/3D presentation of simulated data can be interactively interpreted by the surgeon. During the usability tests, the lowest time needed from surgeons was measured for the determination of the configuration with highest overall fixation force. This can be traced to table representation, which gives a good overview of all configurations, and thus, the highest value can be determined without further interaction. The most time needed was measured for identifying the configuration with the best proximal sealing potential. This is due to the high number of necessary interactions, as evaluation has to be performed by the surgeon’s visual interpretation of the 3D model. In the middle is the time needed for determination of the configuration with highest proximal fixation force. The configuration has to be selected from the combo box element before it is possible to evaluate proximal and distal force values. Therefore, more time is needed than to select the overall fixation force from a table. Compared with determination of sealing potential which has to be done solely visual, it needs less time because of the listing of force values. The box plots in Fig. 12 enables a good recognition that with increasing interaction for data interpretation not only the time needed increases but also the distribution of time measured for several users. Due to this,

123

Int J CARS

the user interface for data interpretation should require user interaction as little as possible. For a faster recognition of the information Where is an endoleak type I? one surgeon suggests to mark endoleaks in the color legend instead of good sealing. This results in a visualization of all other ring parts with another color (also the inner ring surfaces which have no contact to other elements). In the context of sealing state evaluation, a few surgeons mentioned the evaluation value contact pressure. The contact pressure is not useful to visually evaluate the sealing because the numerous color-coded values, for example from color blue (0 MPa) over green (0.15 MPa) to red (0.29 MPa) as used in the work of Scherer et al. [15], make it difficult to locate leaky areas. A binary-coded contact state with values contact exists or no contact seems to be clearer. Furthermore, analog to fixation force, it is not yet known what contact pressure could be defined as a sufficient threshold value for ensuring a tight fit over long-term course. The interactive work of the surgeons with simulation results shows that the originally desired visualization of force distribution on the stent ring [17] can not be interpreted by most surgeons; moreover, the color legend is confusing when working with the module for determination of proximal and distal fixation forces. Some surgeons thought a homogenous force distribution over the ring could lead to a better fixation. When analyzing the force sum per ring using bar charts, a homogenous force distribution over all rings was interpreted by surgeons as a good result. This is an interesting starting point for further work to examine whether a homogenous force distribution leads to better fixation than inhomogenous force distribution in long-term behavior. Generally, it has to be investigated whether it is possible to determine a force value per ring which enables classification of rings regarding their contribution to stent graft fixation within the flow field. Some of the respondents consider the module for evaluation of ring forces too complex and would not use it for optimization of leg length. For this reason, it should be examined exactly how far such a module can assist the surgeon in stent graft selection and also deletion of this module should be considered. During the usability test, most surgeons asked for the meaning of overall fixation force. The crucial factor to integrate this value in the analysis module is the data for stent graft drag forces in the literature that are used to compare different stent graft devices. Interviews with surgeons show that an overall fixation force can not be evaluated by them because there is no reference value known. Therefore, the addition of all forces provides no added value. Proximal and distal fixation force can be different, which is not recognizable if only the overall forces are indicated. For example, a stent graft with highest overall fixation force can have high distal but insufficient proximal fixation forces, and therefore will probably lead to a stent graft migration. Further work

123

should determine a configuration proposal based on results of the module Migration risk. The configuration with sufficient fixation forces in all landing zones should be the recommended configuration. If there are multiple configurations that fulfill this precondition, the one with highest proximal forces has to be proposed, because this value was rated as more important than distal forces by most surgeons. Visualization methods implemented for 2D/3D data presentation meet the aesthetics and intuitive interpretability required by the surgeons. Interaction elements provided enable an explicit data access and could be managed by the surgeon on his own after a short introduction. It was found, however, that a reduction of four interaction modules (configuration proposal, evaluation single configuration, comparison of multiple configurations and analysis migration risk) to two modules is desirable. Whereby repeated jumping back and forth can be avoided. Furthermore, it has to be considered that there is not excessive information presented to the surgeon that cannot be evaluated, for example indication of forces in Newton, which underly the final processed results. The majority of evaluators consider the module for comparison of multiple stent grafts as helpful as there are often only two to three stent graft systems used in clinical routine. Therefore, other products can not be compared due to missing product knowledge and experience reports. Besides stent graft properties, a safe handling by the surgeon is crucial to therapy quality. Device implantation is improved by frequent and regular handling; therefore, many surgeons would not use the Medical Postprocessor for stent graft comparison when treating standard anatomies. However, an additional expenditure of time for quantitative and qualitative evaluation of multiple devices and configurations seems to be warranted for treatment of complex aneurysm morphologies, for example those with short overall length (from renal artery outlet of the aorta to bifurcation), heavy aneurysm neck angulation, or short iliaca communis (Incidence: approximately 20 % of all cases). The off-label stenting is a new trend which treats heavy aneurysm neck angulations with products that are not approved for such angulations. Here, the simulation results can show which stent graft product has the lowest migration and endoleak type I risk. Furthermore, the simulationbased stent graft planning system can assist when the surgeon decides to use a “hybrid stent graft system” which combines components from different manufacturers. One more example how the analysis of the suggested stent graft properties can influence the clinical decision is the determination of the oversizing of the stent graft components. So far, this has been done using a thumb rule of 10–20 % oversizing. Is there, for example, a vessel with a diameter of 23 mm and the surgeon wants to use the Anaconda stent graft system [6], he has to decide between a bifurcation diameter of 25 mm, which is approximately 8.6 % oversizing or the next higher size with 28 mm, which means an oversizing of

Int J CARS

approximately 21.7 %. In this case, the simulated stent graft properties for each size may give the surgeon information about which bifurcation diameter provides a better sealing and has lower migration risk. The above-mentioned examples how the simulation-based approach may assist stent graft selection are based on the feedback from vascular surgeons gained during the usability test for the presented Medical Postprocessor. However, to finally determine the benefit of this method in clinical practice, a study has to be performed perspectively which compares the conventionally made selection of a stent graft by a vascular surgeon with the selection done by using the simulation-based planning system. Are they different, it has to be discussed with the surgeon how far the analyzed stent graft properties influenced his decision process. Furthermore, the stent graft proposal of the planning system can be checked against the surgeon’s selection. The evaluation of the user interface certifies a good usability, which can be traced back to the usage of common interaction elements like tabs, combo boxes and icons for interaction elements. The criterion individualization has the fewest points due to a missing feature for individual adjustments of specific software components. A limitation of the Medical Postprocessor presented is the non-automated evaluation of calculated contact states. The reason for this is the meshing of stent graft rings in the FE software: after the simulation, all stent graft elements have a contact state, even elements on the inner side of the ring. Due to mesh definition, a contact state with value 1 for all elements (both inside and outside) is not possible. Therefore, a query for contact state with value 1 (contact exists) for all elements is not reasonable. Which contact elements of the stent graft are in contact with surrounding tissue after structural simulation (expansion of stent graft) cannot be predicted. Thus, a localization of eligible elements or determination of the amount of potential elements is not possible. A further difficulty is the requirement of most vascular surgeons for continuous sealing over multiple rings for a minimum total length of 10 mm. Therefore, it must additionally be examined whether rings with complete sealing occur sequentially for the required length. Due to the described problem, an automated evaluation of contact states is currently not possible. Regarding the visualization of contact states on stent graft rings, one must bear in mind that inner surfaces of the rings have also a white coloring because they have no contact with further model parts.

Conclusion A Medical Postprocessor for interpretation of FE results was introduced, which may lead to a better estimation of stent graft fixation and sealing potential in the context with

patient-specific vessel. On the basis of fixation forces and contact state, a comparison of multiple devices and configurations is possible. The gains of the Medical Postprocessor are not expected to be in faster planning times, but rather in an increased quality of therapy due to the personalization of the device selection to patient vessel morphology. The slight increase in planning time is a tradeoff for a predicted higher rate of long-term implant success due to the use of quantitative and qualitative, patient-specific data. 2D and 3D data presentation forms and suitable interaction methods were implemented for enabling quantitative and qualitative data analysis. The features and user interface developed for the Medical Postprocessor were evaluated by multiple vascular surgeons. The proposed stent graft properties for stent graft evaluation, as well as data presentation and interaction methods for assisting stent graft selection were evaluated by these surgeons as helpful. Assessment of the software with regard to ISO Standard 9241-110 principles results in an user-friendly user interface without requirement for improvement. The result of the usability tests show that the Medical Postprocessor can enable a more in-depth examination of available products in context with patient-specific vessel directly by the surgeon. The Medical Postprocessor serves as an example of how quantitative FE methods can contribute to personalized medicine and shows the promise of emerging role of simulations in qualitative and quantitative stent graft evaluation. The feedback from surgeons who interacted with the 3D result model identified further important model parameters. Besides calcified plaque, modeling of thrombus in landing zones is important, as both can become deformed due to pressure caused by stent graft expansion and thereby induce a modification of vessel lumen. Furthermore, thrombus deformation can cause an occlusion of neighboring vessels. Stent graft coating for simulation of device flexibility in the vessel lumen should also be considered for a more precise length determination of the iliac leg. The integration of blood pulsation and its impact on stent graft radial forces is very important and demanded from the surgeon’s side. In this context, a simulation for long-term behavior with multiple flow cycles should also be implemented. Then, potential progressive changes in postoperative aneurysm neck can be better estimated. The Medical Postprocessor gives vascular surgeons a deeper insight and better understanding of the potential use of finite element models for EVAR planning. Furthermore, model extensions and additional model results can be determined by interactive feedback development cycles with surgeons. The Medical Postprocessor introduced can be used in further research work as a discussion and development platform for determining an intervention-specific simulation model.

123

Int J CARS Acknowledgments This work was sponsored by funds of the European Regional Development Fund (ERDF) and the state of Saxony within the framework of measures supporting the technology sector.

16.

17.

Conflict of interest Sandra von Sachsen, Björn Senf, Oliver Burgert, Jürgen Meixensberger, Hans-Joachim Florek, Friedrich Wilhelm Mohr and Christian Dirk Etz declare that they have no conflict of interest.

18.

References

20.

1. Volodos NL, Shekhanin VE, Karpovich IP, Troian VI, Gur’ev IA (1986) A self-fixing synthetic blood vessel endoprosthesis. Vestn Khir Im I I Grek 137:123–125 2. Parodi JC, Palmaz JC, Barone HD (1991) Transfemoral intraluminal graft implantation for abdominal aortic aneurysms. Ann Vasc Surg 5:491–499 3. Dake MD, Miller DC, Semba CP (1994) Transluminal placement of endovascular stent-grafts for the treatment of descending thoracic aortic aneurysms. N Engl J Med 331(26):1729–1734 4. Stelter WJ, Umscheid T, Ziegler P (1997) Three-year experience with modular stent graft devices for endovascular AAA treatment. J Endovasc surg 4:362–369 5. Deutsche Gesellschaft für Gefäßchirurgie (DGG) (2008) 8. Qualitätsmanagement Bauchaortenaneurysma 6. VASCUTEK Ltd, a TERUMO Company, Anaconda stent graft, http://www.vascutek.com/vascutek/products/item/anaconda-aaastent-graft-system. Active link, July 2013 7. Kortmann H (2011) Bauchaortenaneurysmen (Kapitel 11). In: Kompaktwissen Gefäßchirurgie, B.L.P. Luther (Hrsg.). Springer, Heidelberg, pp 199–219 8. Rosset A (2012) OsiriX imaging software. http://www.osirixviewer.com. Active link, July 2013 9. Aquarius iNtuition Enterprise (2012) TeraRecon. http://www. terarecon.com. Active link, July 2013 10. Raghavan ML, Vorp DA (2000) Toward a biomechanical tool to evaluate rupture potential of abdominal aortic aneurysm: identification of a finite strain constitutive model and valuation of its applicability. J Biomech 33:475–482 11. Vascops GmbH (Österreich), VASCPOS (Sweden). http://www. vascops.com/en/vascops-A4clinics.html. Active link July 2013 12. Molony DS, Broderick S, Callanan A, McGloughlin TM, Walsh MT (2011) Fluid-structure interaction in healthy, diseased and endovascularly treated abdominal aortic aneurysms. Stud Mech Anobiol Tissue Eng Biomater, vol 7, Springer, Berlin, pp. 163– 179. doi:10.1007/8415_2011_85. ISBN:978-3-642-18094-1 13. Sun Z, Chaichana T, Sangworasil M, Tungjitkusolmun S (2008) Computational fluid analysis of blood flow characteristics in abdominal aortic aneurysms treated with suprarenal endovascular grafts. ICBME Proc 23:1728–1732 14. Figueroa CA, Taylor CA, Yeh V, Chiou AJ, Zarins CK (2009) Effect of curvature on displacement forces acting on aortic endografts: a 3-dimensional computational analysis. J Endovasc Ther 16:284– 294 15. Scherer S, Treichel T, Ritter N, Triebel G, Drossel, WG, Burgert O (2011) Surgical stent planning—simulation parameter study for

123

19.

21.

22.

23.

24.

25.

26. 27. 28. 29. 30.

31.

32.

33.

34.

models based on DICOM standards. Int J CARS, 6, 3. Springer, Heidelberg, pp 319–327 de Bock S, de Beule M, de Santis G, Vermassen F, Segers P, Verhegghe B (2011) A simulation tool for virtual stent graft deployment in patient-specific abdominal aortic aneurysms. In: Proceedings of the ASME, summer engineering conference von Sachsen S, Etz CD, Mohr FW, Senf B, Neugebauer R, Florek HJ (2011) A method for integrating finite element results in a surgical planning software for evaluating stent graft properties in endovascular surgery. In: IEEE proceedings 4th international conference on biomedical engineering and informatics, Shanghai (China), pp 1349–1354. ISBN:978-1-4244-9350-0 Gebert de Uhlenbrock (2009) A. Dissertation Technische Universität Hamburg-Harburg, Designanalyse von endovaskulären Aortenprothesen Senf B, von Sachsen S, Neugebauer R, Drossel WG, Florek HJ, Mohr FW, Etz CD (2013) The effect of stent graft oversizing on radial forces considering Nitinol wire behavior and vessel characteristics. J Med Eng (accepted) Sinha Roy A, Westt K, Rontala RS, Greenberg RK, Banerjee RK (2007) In vitro measurement and calculation of drag force on iliac limb stentgraft in a compliant arterial wall model. Mol Cell Biomech 4(4):211–226 Corbett TJ, Molony DS, Callanan A, Mc Gloughlin TM (2011) The effect of vessel material properties and pulsatile wall motion on the fixation of a proximal stent of an endovscular graft. Med Eng Phys 33(1):106–111 Arko FR, Heikkinen M, Lee ES, Bass A, Alsac JM, Zarins CK (2004) Iliac fixation length and resistance to in-vivo stentgraft displacement. In: Proceedings of the 19. Annual meeting of the western vascular society, vol 41, number 4, pp 664–671. doi:10.1016/ j.jvs.2004.12.050 Kleinstreuer C, Li Z, Basciano CA, Seelecke S, Farber MA (2008) Computational mechanics of Nitinol stent grafts. J Biomech 41:2370–2378 Gao F, Watanabem M, Matsuzawa T (2006) Stress analysis in a layered aortic arch model under pulsatile blood flow. BioMed Eng Online 5. doi:10.1186/1475-925X-5-25 Holzapfel GA, Sommer G, Regitnig P (2004) Anisotropic mechanical properties of tissue components in human atherosclerotic plaques. J Biomech Eng H 126:657–665. doi:10.1115/1.1800557 Materialise (2013) Mimics innovation suite. http://biomedical. materialise.com/mimics. Active link July 2013 Materialise (2013) 3-matic. http://biomedical.materialise.com/3matic-0. Active link July 2013 ANSYS, Inc. (2009) ANSYS manual; Version 11 Kitware (2013) The visualization toolkit. http://www.vtk.org. Active link July 2013 German Cancer Research Center Division of Medical and Biological Informatics Heidelberg (2012) Medical interaction and imaging tool kit (MITK). http://www.mitk.org/wiki/BlueBerry. Active link July 2013 Murphy EH, Johnson D, Arko FR (2007) Device-specific resistance to in vivo displacement of stent-grafts implanted with maximum iliac fixation. J Endovasc Ther 14:585–592 Bosman WMPF, Steenhoven vd, Steenhoven TJ, Suárez DR, Hinnen JW, Valstar ER, Hamming JF (2010) The proximal fixation strength of modern EVAR grafts in a short aneurysm neck. An in vitro study. Eur J Vasc Endovasc Surg 39:187–192. doi:10.1016/j. ejvs2009.10.019 ISONORM 9241-110 (2013) Ergonomics of human-system interaction—Part 110: dialogue principles. http://www.iso.org/ iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=38 009. Active Link July 2013 Prümper J (1997) Der Benutzungsfragebogen ISONORM 9241/10: Ergebnisse zur Reliabilität und Validität. In Liskkowsky R,

Int J CARS Velichkovsky BM, Wünsschmann W (Hrsg) Software-Ergonomie ’97-usability engineering: integration von Mensch-ComputerInteraktion und Software-Entwicklung, pp 253–262 35. Prümper J (2013) ISONORM 9241-110 questionnaire. http://www. seikumu.de/de/dok/dok-echtbetrieb/Fragebogen-ISONORM-924 1-110-S.pdf. Active link July 2013

36. Prümper J (2013) ISONORM 9241-110 Questionnaire interpretation matrix. http://www.seikumu.de/de/dok/dok-echtbetrieb/ Vorlage-Auswertung-ISONORM-Fragebogen.xls. Active link July 2013

123