Recent advances in cerebrovascular simulation and neuronavigation for the optimization of intracranial aneurysm clipping

Abstract

Endovascular treatment of intracranial aneurysms (IAs) has improved to the extent that in some instances such an approach has now become safer than surgery. This has dramatically changed clinical practice by reducing the volume and increasing the complexity of IAs referred for open surgical treatment. We review the simulation techniques and dedicated vascular neuronavigation systems that have been developed to maintain the quality of aneurysm clipping in this context. Simulation of surgical approaches was made possible by the introduction of high-resolution 3D imaging techniques such as three-dimensional CT angiography (3D-CTA) and three-dimensional digital subtraction angiography (3D-DSA), enabling reproduction of the craniotomy and rotation of the vascular tree according to the orientation of the operative microscope. A virtual simulator for compiling such data, the Dextroscope®, is now available for this purpose. Simulation of final clipping has been investigated through virtual or physical models, enabling anticipation of aneurysm deformation during clip application and selection of the appropriate clip(s) in terms of number, size, shape and orientation. To improve surgical dissection guidance, specific cerebrovascular neuronavigation procedures have been developed based on 3D-CTA or 3D-DSA. These help make the operation secure by accurately predicting the location and orientation of an aneurysm within its parenchymal and vascular environment. Future simulators dedicated to cerebrovascular procedures will need to integrate representation of the brain surface and biomechanical modeling of brain and aneurysm wall deformation under retraction or during clipping. They should contribute to training and maintenance of surgical skills, thereby optimizing the quality of surgical treatment in this field.

• Simulation
• intracranial aneurysm
• clipping
• image guided surgery
• neuronavigation
• surgical planning

Introduction

The prevalence of intracranial aneurysms (IAs) varies between 0.5% and 6% in the general population. Rupture of these vascular malformations usually occurs at the weakest point in the aneurysm wall. This is a very serious complication resulting in subarachnoid hemorrhage, which is still associated with a very uncertain prognosis (morbi-mortality of 40–45%) [1]. When such aneurysms are discovered incidentally, the aim of management is to select for treatment those patients having the highest risk of aneurysm rupture. In both situations (i.e., ruptured aneurysms or incidental discoveries), the safest and more efficient procedure should be proposed, whether endovascular treatment (i.e., coiling with or without balloon remodeling or stenting) or microsurgical clipping.

The ISAT (International Subarachnoid Aneurysm Trial) is the only multicenter international and randomized study (2143 patients) that compared these two treatment options for the management of ruptured IAs [2]. It showed that the mortality or dependence rate at one year was lower in the endovascular group (23.7%) than in the surgical group (30.7%). Although late analysis of the results suggested that differences between the two treatments might not be significant on long-term follow-up, this study considerably changed clinical practice all over the world and resulted in an increase in the number of aneurysms treated through endovascular means. In this context, the number of aneurysms referred for microsurgical clipping dramatically decreased at many centers. Some studies have shown that in-hospital mortality and morbidity after surgical treatment of IAs is lower in high-volume hospitals than in low-volume institutions [3], [4]. Thus, the reduction in the number of operated cases could affect the level of neurosurgical skill in this field. At the same time, surgical clipping has often become a second-line treatment proposed for more complex IAs deemed unsuitable for embolization.

In the context of sub-specialty training and development of high-volume centers, and to address the goal of maintaining good surgical results, neurosurgeons have been led to develop new training methods, specific surgical planning procedures, and dedicated neuronavigation-guided vascular approaches. All of these solutions are based on modern imaging techniques such as multiple-detector three-dimensional CT angiography (3D-CTA), three-dimensional digital subtraction angiography (3D-DSA), or magnetic resonance imaging (MRI) or angiography (MRA). The combination of 3D reconstructions from these imaging modalities allows neurosurgeons to analyze precisely the vascular anatomy and aneurysm morphology, and to anticipate anatomical difficulties that they will encounter during surgery [5–25].

Malone et al. [26] have provided a general review of recent methods for neurosurgical assistance with computer-based simulation environments (Dextroscope®, cranial base surgery simulators, ROBO-SIM®, ImmersiveTouch®, etc.). In the present review, we focus specifically on the available techniques that allow simulation of IA clipping or cerebrovascular neuronavigation (Table I).

Table I.  Methods of simulation and neuronavigation for intracranial aneurysm clipping: characteristics and applications

Techniques	Suitable for emergency procedures	Planning purposes	Educational purposes	Simulation including bone environment	Simulation including cerebral environment	Simulation of clip application	Cost
3D-imaging alone [20], [24], [30]	+	+	+	+/−	−	+/−	−
Dextroscope [31], [32]	−	+	+	+	−	+	+++
Virtual model [33–35]	−	+	−	−	−	+	+
Physical model [36], [37]	−	+	+	+/−	−	+	++
Neuronavigation guidance [39–45]	+/−	+	−	+	−	−	+

Clipping simulation with 3D-CTA and 3D-DSA

Various factors affect the success of IA surgery. One of them is the ability of the neurosurgeon to mentally or visually anticipate potential difficulties in dissecting and clipping the aneurysm. Three-dimensional radiological techniques, i.e., 3D-CTA [27], 3D-MRA [28] and 3D-DSA [29], are useful tools for analyzing the arterial relationships and projection of IAs, and thereby predicting the pitfalls and risks of surgery.

Recently, Siablis et al. [20] reported the use of a 3D-CTA protocol that allows reproduction of surgical views of anterior circulation aneurysms. Our group also developed a semi-automatic 3D-DSA protocol displaying direct views of the arterial tree oriented according to the fronto-pterional approach, and validated this algorithm through correlations with intraoperative views [24]. The rotational angles were approximated by observation of the patient's head position and microscope axis orientation during the surgical procedure. The specific angles for each aneurysm location were then determined. A coronal angle of 60° was suitable for representation of the exposition of anterior communicating aneurysms; in other locations a 45° rotational angle was more appropriate. Obviously, these angles did not summarize the multiple positions of the microscope used for this surgery, and the reconstructions could not entirely replace the precise analysis of the 3D data from multiple angles on the imaging workstation. Nevertheless, they provided a global and direct view of the vascular conformation through a fronto-pterional approach, as would be seen by the neurosurgeon after exposure of the anterior circulation arteries and aneurysm. Preoperatively, this protocol facilitates the mental representation of the 3D arterial relationships of the aneurysm and enables potential difficulties in the clipping procedure to be predicted. Peroperatively, the protocol enables the surgeon to avoid beginning the dissection at the site of the aneurismal bleb, since it shows the exact 3D orientation of the aneurismal sac and depicts the precise location of neighboring branches (particularly those beside the aneurismal sac), allowing adjustment of the clip application. A drawback of this technique is that it does not represent surrounding structures such as brain parenchyma and nerves, which must be taken into account when planning this kind of procedure.

In 2004, Futami et al. [30] tried to estimate the risk of a postoperative aneurysm remnant by simulating aneurysm clip application on 3D-CTA reconstructions. In this study, 36 IAs (≤10 mm in size) were evaluated. All the structures around the aneurysm except bone and collateral arteries were removed. “Operative views” were obtained by rotating the 3D volume according to the chosen surgical approach. A linear image subtraction representing the thickness of a straight aneurysm clip was applied with ideal orientation at the site where it was planned to clip the aneurysm. Image analysis after clipping “simulation” enabled identification of potential aneurysm remnant below the virtual clip projection. Two neurosurgeons performed this preoperative evaluation independently. The final position and orientation of the clip were recorded if both observers reached the same conclusion independently or by consensual agreement. Finally, all the patients were operated by a third neurosurgeon, and all the IAs were excluded with a similar titanium straight clip. Intraoperative endoscopy was performed to evaluate the presence of aneurysm remnant, and if necessary the clip was repositioned. The quality of aneurysm obliteration was assessed by conventional postoperative angiography. This confirmed complete obliteration of 19 out of 21 aneurysms initially expected to become totally excluded according to the preoperative simulation, and found aneurysm remnants in 9 out of 15 expected cases. In this inaugural study, the authors used the cut-along-trace function of their 3D-CTA workstation to represent the clip position and orientation in order to predict postoperative aneurysm remnants. This simulation of clip application was rather primitive, but represented a first step towards more complex representations. In addition, they did not correlate the clipping simulation with the intraoperative findings in terms of clip position and remnant size or location. Neither did they represent the aneurismal environment and the potential difficulties of placing a clip through a surgical corridor narrowed by surrounding brain structures.

Clipping simulation with the Dextroscope®

The Dextroscope® (Volume Interactions, Singapore) is a virtual reality environment that enables planning and training for operative neurosurgical procedures. Patient-specific radiological data (cerebral MR and CTA) are loaded, colored and highlighted on the Dextroscope® workstation, and a virtual 3D patient model is then created. The stereoscopic image is displayed on a monitor and reflected into the user's eyes by means of a mirror. Users have a virtual stylus in their hand that is calibrated to transmit their movements to their tools in the virtual space [31].

For use with this simulator, Wong et al. [32] designed an operative application for training in craniotomy and aneurysm clipping. They loaded onto the workstation patients’ 3D-CTA and cranial bone 3D CT scans, as well as digitized clips and clip applicators. The drill function of the software was used to perform a craniotomy and thus expose the vascular tree and the aneurysm according to the expected operative view. The user was then able to select the appropriate clip and simulate the clipping by advancing the virtual clip along the aneurysm neck. This tool allowed modeling of the ideal craniotomy to obtain the optimal and most realistic surgical exposure of the aneurysm, but also enabled the user to choose the most suitable clip (in terms of shape, size and orientation). Despite all these advantages, however, the described protocol does not represent the aneurismal environment (veins, brain parenchyma, nerves, etc.). Hence, it only provides a partial representation of the surgical approach and of the potential difficulties involved in reaching, correctly exposing and treating IAs. Furthermore, while the 3D rendering of the vascular tree is very appealing, this technique lacks haptic operator controls to transmit tissue strengths and clip applicator force feedback to the user. Finally, the main disadvantage of the Dextroscope® is its very high cost, which makes this simulator hardly affordable for a single institution.

Clipping simulation using virtual models

Koyama et al. [33–35] created an application program that represented the properties of a saccular basilar tip aneurysm. They then digitized clips and clip applicators and simulated the clip movements. The morphological changes caused by clip compression at the aneurysm neck (branching site) were also simulated by a sine- and cosine-based formula. Finally, they created a computer simulation program in which this aneurysm was clipped via a right trans-sylvian approach through the basal cisterns and perforating arteries.

This study enabled an important breakthrough in the simulation of IA morphological response to manipulation during clipping. The protocol could be very useful in predicting aneurysm deformation under forces applied by the clip jaws at the neck. This information is of great importance when evaluating the ideal size, shape and orientation of the definitive clip. However, the clinical reality might be more complex because patient-specific biomechanical properties of the aneurysm wall are poorly documented. The properties of a thin, smooth aneurysm wall could be very different from those of a thick atherosclerous or calcified sac. Depending on their variable histological structure, IAs might not react to the positioning of the clip in the exact same way as in preoperative simulations. Further studies are needed to better characterize patient-specific aneurysm wall properties and to improve the reliability of these simulations. Another important aspect that was not addressed by this study is the impact of clip application on vessel morphology. Indeed, classical IAs arise at arterial bifurcations (branching site aneurysms), and obliteration of the neck might induce modifications of parent and collateral branches. Sometimes, depending on the clip orientation (if perpendicular to the bifurcation axis), shape or position at the neck (if close to the bifurcation), clip application could lead to arterial deformation or kinking, and result in secondary arterial occlusion and stroke.

Clipping simulation using solid or elastic physical models

D’Urso et al. [36] were the first to replicate the cerebral vasculature morphology of patients in a solid material. Nineteen biomodels were obtained by downloading native images from CTA and MRA onto a dedicated computer workstation. Raw image data were then converted with the Anatomics BioBuild® biomodeling system (Anatomics, Melbourne, Australia) to a format compatible with a stereolithography apparatus for manufacturing the models. Initial 3D reconstructions were performed using a volume rendering technique. The segmentation between vessels and bone was achieved by image thresholding and structures unconnected to the main arterial tree were removed using a 3D connectivity function. Trilinear gray-value interpolation was applied to generate the contours of each slice and between contiguous slices. The contour data were then used to create the final object file that was sent to the stereolithography device. In the manufacturing process, a laser beam solidified layers of a photosensitive liquid resin monomer according to the cerebral vasculature contours. The resulting object was then rigidified in an ultraviolet (U.V) oven. The utility of biomodels was subjectively assessed by the neurosurgeons in charge after the operation. They reported that the biomodels accurately represented the cerebral vasculature and the aneurysm relationships except in one case (an endosaccular thrombus). Thus, this tactile anatomic overview would help even an inexperienced surgeon to quickly understand the spatial organization of the aneurysm without requiring complex mental reconstruction from multiple images or replacement of the vascular volume. The models helped optimize the position of the patient's head with respect to the most appropriate approach angle, and were also helpful for understanding the 3D anatomy, which gave the surgeon more confidence during the procedure. It was also possible to try the appropriate clip in terms of length, shape and orientation on the model, developing a de facto new type of direct simulation.

Kimura et al. [37] made 3D elastic hollow models of individual cerebral aneurysms for the purpose of preoperative simulation and surgical training (3 retrospective and 7 prospective cases). They also applied a stereolithographic technique and used a prototyping machine to build the model from a rubber-like polymer hardened under u.v. light, according to the vessel wall anatomy. The aneurysm model was then fixed with either flexible wires or plastic clay, according to the selected approach, and oriented along the surgical view. Finally, under the operating microscope, various types of aneurysm clip were applied to determine the most appropriate size, shape and orientation. In one case of a deep-seated vertebrobasilar aneurysm, they designed a solid 3D model including the aneurysm, vessels, and cranial base bone. They then created a craniotomy and simulated the access to the aneurysm. The goal of developing these models was to represent a real 3D arterial tree in order to simulate preoperatively the surgical repair of IAs (with regard to selection of clip properties and orientation). This technique might also assist young neurosurgeons in developing their own surgical strategy and allow them to confront the potential difficulties of an approach and clip application in a narrow corridor.

Unfortunately, these simulation techniques lack any representation of surrounding brain structures, which is one of the main aspects restricting accessibility and maneuverability of the aneurysm. When looking at anatomic accuracy, the authors acknowledged difficulties in replicating small collateral arteries and avoiding contamination of the solid biomodels by venous components that could only be distinguished from the arterial component by an experienced vascular neurosurgeon. In two cases using solid biomodels, the aneurysm neck was also poorly depicted because of the limited definition of native images combined with suboptimal segmentation of the vasculature (operator-dependent). Another drawback of these methods is the absence of information concerning the thickness or biomechanical properties of the aneurysm wall and parent vessels, which could be useful in predicting their deformability during clipping. Finally, these interesting techniques are quite expensive, and the preparation of a single model takes usually several days, making such simulation solutions hardly applicable to emergency situations like ruptured IAs.

Neuronavigation-guided aneurysm clipping

To improve the accuracy of neurosurgical procedures, especially in cerebral tumor surgery, neuronavigation systems are now commonly used. These image-based localization systems include a workstation and a reference array or frame with passive markers which reflect infrared flashes and are detected by an infrared camera system. The reference array or frame is usually connected to a head-holder clamp immobilizing the head. Before the operation, imaging data are loaded onto the workstation and a 3D reconstruction of the patient's head is obtained. Patient-to-image registration is then performed precisely in the operating room by matching several fixed reference points (i.e., reproducible anatomical landmarks of the face or fiducials positioned preoperatively on the scalp surface) that are identified on both the 3D image and the surface of the patient's head. Once this correlation is validated, it is possible to follow the progress of the surgical approach by pointing with a probe to a region of interest in the operative field that will be instantly displayed on the workstation screen [38]. Recent improvements in the technique enable calibration of the surgical microscope as a pointer. In this surgical microscope guidance application, the focal point represents the tip of a virtual probe and is used for neuronavigation.

Several groups have applied this technique to IA surgery using 3D-CTA data [39–42]. The datasets obtained were loaded onto the neuronavigation system (Medtronic StealthStation® or Brainlab VectorVision®) to construct 3D volume-rendered models of the vessels using dedicated 3D software implemented on the workstation. The 3D vessel models were displayed on the screen and enabled location of the aneurysm and its surrounding parent vessels at every step of the surgical approach. This technique has the advantage of providing automatic and fast post-processing of native images into a 3D vascular model. The application was rated as being very helpful in making the surgical procedure secure by predicting the precise position and orientation of the aneurysm sac, and particularly its rupture site. The method was also used to fashion less invasive surgical approaches through which the aneurysm could be reached and, to a certain extent, clipped “blindly” (i.e., without controlling its back wall). Moreover, the neuronavigation system could be connected to the operative microscope and transparent computer-generated reconstructions of the vascular tree could be displayed in the microscope eyepieces. This augmented reality technology, called a “head-up display” (HUD) (commonly used by air force pilots), allowed guidance of the surgical approach without the surgeon having to leave the operative field in order to follow the progress on the workstation. The technique has already been applied to cerebral tumor resection in areas around eloquent cortex [43]. The authors of this latter study have also suggested the potential usefulness of this technique for quick location and emergency clipping of the aneurysm neck in cases of ultra-early rupture, which is the most unfavorable situation in IA surgery.

Some groups have managed to integrate 3D rotational angiography into a surgical neuronavigation system which is usually only compatible with CT or MR imaging modalities. Indeed, 3D-DSA data do not contain any surface markers of the patient's anatomy to allow registration between the 3D reconstruction and the patient's head. To solve this problem, Raabe et al. [44] developed a method using the coordinates from the angiography system. A special three-point head frame, with three radiographically visible markers at defined distances (two at the external auditory canals and one at the nasion), was used before and during 3D-DSA acquisitions to record the position of the patient's head with respect to the coordinates of the angiography system. The same head frame was repositioned in the operating room to calibrate the 3D-DSA reconstruction loaded onto the workstation (Brainlab VectorVision 2®) with the patient's head. In this study, a good correlation was demonstrated in all cases between the 16 reoriented 3D-DSA reconstructions and intraoperative vascular anatomy (maximum error: 9°). It was found that 3D-DSA guidance facilitated the surgical procedure in 50% of the cases by providing a more accurate location for the aneurysm and collateral arteries that were sometimes covered by brain parenchyma or blood clots (ruptured aneurysms).

Willems et al. [45] developed a procedure enabling direct neuronavigation in the angiography suite. A specific software module was used to determine the positional relationship between the 3D vascular volume and a tracker plate on the 3D-DSA image intensifier through a calibration phantom. The relationship between the tracker plate and the image coordinates was then calculated and stored on the workstation. The positional relationship between the patient's head and the tracker plate was registered. Following 3D-DSA acquisition, a probe under navigation (Medtronic StealthStation Treon®) was used to determine and register the position of superficial fiducials (outside the angiogram volume) relative to the 3D-DSA volume. The final data set was sent to the neuronavigation workstation in the operating room, where the patient-to-image registration and navigation were performed conventionally. The authors were able to visualize the 3D-DSA through the HUD system of their surgical microscope. The main advantage of this method is the use of the gold-standard imaging technique to improve the guidance and precision of the surgical procedure.

The main limitation of these two applications of 3D-DSA for aneurysm surgery guidance is related to the selective catheterism of intracranial arteries during angiography. This might be responsible for the absence on reconstructions of one of the major arteries, despite its direct relationship with the aneurysm, which could be particularly disadvantageous in midline aneurysms filed bilaterally (i.e., anterior communicating aneurysms). Furthermore, 3D-DSA guidance allowed only for the continuous reorientation of the arterial tree according to the surgical view. Indeed, the methodology (perspective registration) and systematic errors during registration precluded direct navigation and accurate location of an intracranial landmark peroperatively.

Future improvements and applications

The development of high-resolution 3D imaging techniques has influenced considerably the way IAs are managed nowadays. These techniques have dramatically improved the understanding of the angioarchitecture of the aneurismal sac and its direct relationships (i.e., with collateral arteries, brain parenchyma and cranial nerves). This is of considerable importance for the safety of the clipping procedure, the goals of which are to obtain complete occlusion of the aneurysm while maintaining the patency of parent and direct collateral arteries as well as preserving neurological function. To accomplish this, the neurosurgeon must first choose the surgical approach that will provide the optimal exposure of the arterial tree and aneurysm while facilitating the exclusion without endangering important brain structures: aspects to be considered include the side (for midline IAs), head orientation, bony opening and removal, brain relaxation techniques, and appropriate opening of arachnoid cisterns along the vessels. The second step is to safely reach the parent artery and the aneurysm while circumventing the usual pitfalls that could lead to suboptimal clip application or neurological sequelae: excessive brain retraction, uncontrolled ultra-early or premature aneurysm rupture, and prolonged temporary occlusion of parent arteries to secure the clipping or to stop peroperative bleeding. The third step is to carefully dissect the aneurysm neck (branching site) and sac from the safest to the weakest part (blebs, rupture site) so as to avoid or otherwise hinder peroperative aneurysm rupture. This has to be pursued sufficiently to visualize the different sides of the aneurysm wall and to identify the collateral branches. The final step is to apply one or several clips at the neck to obtain complete aneurysm occlusion. The quality of this ultimate phase is very dependent on the previous stages (the surgical corridor allowing the introduction of the clip and its applicator in the appropriate orientation, circumferential exposure of the neck, and satisfactory understanding of the aneurysm limits and identification of collateral arteries), but also on the surgeon's anticipation of aneurysm deformability during clip reconstruction according to the bifurcation anatomy and vessel wall quality.

Thus, the success of aneurysm clipping relies on a combination of strict preoperative planning, real-time adapted peroperative strategy and acquired experience. The surgical procedure must be planned using high-resolution multimodality 3D image data. While it is now possible to simulate the surgical exposure (craniotomy, skull base and intracranial angioarchitecture) with 3D-CTA, 3D-DSA or the Dextroscope®, these techniques do not depict the surrounding brain surface, which is one of the main factors restricting accessibility (the surgical corridor) or exposure of the aneurysm, as well as the introduction of clips and their applicators. Future improvements will be to include the brain parenchyma surface in the planning and to simulate its retraction and potential interaction with surgical instruments according to the chosen approach. Because of inter-individual anatomical variability and environmental modifications due to blood clots in ruptured IAs, neuronavigational guidance could be helpful in locating the aneurysm and its direct collaterals. Nevertheless, the dissection is usually performed in a stepwise manner, starting at a reasonable distance from the aneurysm, along the main arterial trunks towards the aneurysm sac. Thus, for an experienced neurosurgeon, it is less important to be guided precisely to the aneurysm or its direct collateral arteries than to understand or visualize their 3D relationships within the parenchymal environment with respect to the surgical approach. This understanding could be obtained with either technique, but has been considerably improved with the use of surgical microscope guidance and HUD applications. Finally, several techniques (virtual and physical models) have been developed to simulate the final step in the procedure, namely aneurysm clipping. Some important points pertaining to this stage would be very interesting to investigate further: not only the ideal number, shape, position and orientation of the clips, but also the optimal approach required for an “ideal” clipping. Therefore, future research will have to implement patient-specific aneurysm wall biomechanics and focus on preoperative simulation of aneurysm wall behavior and collateral branch deformation under clip application in order to predict potential arterial occlusion and optimize clip selection or positioning.

Conclusion

The field of cerebrovascular simulation and neuronavigation is very exciting and is changing rapidly. All the techniques described in this review had direct application to patients with the aim of offering the safest and most optimal procedure in the context of a reduction in the number of IAs being surgically treated while their level of complexity was increasing. Nevertheless, because of their respective limitations and the number of patients that would be required for a statistically meaningful study, the clinical usefulness of these techniques has not yet been demonstrated. Future simulation techniques will have to integrate brain surface representation and biomechanical modeling of brain and aneurysm wall deformation during retraction or clipping, ideally with force feedback reproduction [33], [46–48]. The development of a simulator dedicated to cerebrovascular surgery based on these requirements should certainly help maintain training, skills and expertise in intracranial aneurysm surgery.

Declaration of interest: This work was supported by a research grant from the French Society of Neurosurgery.