Simulation training approaches in intracranial aneurysm surgery—a systematic review

Background With the increasing complexity and decreasing exposure to intracranial aneurysm surgery, training and maintenance of the surgical skills have become challenging. This review elaborated on simulation training for intracranial aneurysm clipping. Methods A systematic review was performed according to the PRISMA guidelines to identify studies on aneurysm clipping training using models and simulators. The primary outcome was the identification of the predominant modes of the simulation process, models, and training methods associated with a microsurgical learning curve. The secondary outcomes included assessments of the validation of such simulators and the learning capability from the use of such simulators. Results Of the 2068 articles screened, 26 studies met the inclusion criteria. The chosen reports used a wide range of simulation approaches including ex vivo methods (n = 6); virtual reality (VR) platforms (n = 11); and static (n = 6) and dynamic (n = 3) 3D-printed aneurysm models (n = 6). The ex vivo training methods have limited availability, VR simulators lack haptics and tactility, while 3D static models lack important microanatomical components and the simulation of blood flow. 3D dynamic models including pulsatile flow are reusable and cost-effective but miss microanatomical components. Conclusions The existing training methods are heterogenous and do not realistically simulate the complete microsurgical workflow. The current simulations lack certain anatomical features and crucial surgical steps. Future research should focus on developing and validating a reusable, cost-effective training platform. No systematic validation method exists for the different training models, so there is a need to build homogenous assessment tools and validate the role of simulation in education and patient safety.


Introduction
In recent decades, the use of noninvasive cerebral vascular imaging has increased, resulting in greater detection of cerebral aneurysms [1,2].Nowadays, these aneurysms are increasingly treated endovascularly [2][3][4], even if certain complex cases are better suited for microsurgical clipping.The number of surgical clipping procedures is reducing [2,5], providing fewer opportunities for neurosurgery trainees to practice clipping.This should be compensated so younger trainees are able to train for aneurysm clipping cases that are unsuited for endovascular treatment.
Although simulated training methods for minimally invasive procedures exist and are available in other specialties like orthopedics [6] and abdominal interventions [7], the simulation of microsurgical aneurysm clipping has only been thoroughly investigated in the last decade.It has been shown that excellent neurovascular training can still be obtained with a dedicated residency program and by improving skills using training models [5].The use of 3D training models improves the understanding of relevant surgical anatomy as they provide knowledge of spatial relationships and a sense of depth [2].Such models help residents to develop basic skills in aneurysm clipping [3,8].With these training models, residents can start practicing clipping procedures at an early stage of their residencies [9].Using simulation models reduces the chance of errors during surgery and refines the surgical strategy, which benefits patient safety [3,8].This study reviews different state-of-the-art approaches for training simulations and compares them with benchtop models, thereby checking if they improve preoperative assistance and medical knowledge among residents.We focused on studies regarding intracranial aneurysm clipping simulation for training neurosurgeons in educational or clinical settings.The aim was to determine the effectiveness and usability of different training methods, as well as the possibility of obtaining hands-on experience.The reported strengths and limitations of each training method were determined along with the fidelity and feasibility of extensive training with the existing methods.The quality of training was also investigated to determine if current training methods can sufficiently prepare practitioners to treat cerebral aneurysms.

Materials and methods
A systematic review was performed to identify research works on intracranial aneurysm clipping training using models and patient simulators.The search adhered to the PRISMA guidelines for reporting systematic reviews [10].

Search strategy
The following electronic databases were searched from their inception until 21 September 2022: MEDLINE, Embase, PubMed, Cochrane, and PsycInfo.Each database search was conducted based on information source logic using Medical Subject Headings (MeSH/EMTREE) and Text Words.The search strategy used for MEDLINE is represented in Table 1; a similar search strategy was used for the other databases.

Inclusion and exclusion criteria
Studies on the development and evaluation of simulation setups for aneurysm clipping training and those that verified the training abilities were considered.Studies based on computer simulations of fluid dynamics and endovascular procedures for aneurysm treatment were excluded.When multiple papers by the same authors and research groups covering the same topic with a similar scope were found, the most recent article was included.To evaluate the true effectiveness of the training approaches, only studies that involved neurosurgeons and/or residents as participants using the model and measured their training outcomes were included.Articles were not excluded based on publication year or language.Only articles on simulations that contributed to a learning process, improved surgical skills, or furthered medical education were included.

Study selection and data extraction
No document restrictions or methodology filters were applied to the primary search.Duplicate records were removed when EndNote 20 software (Clarivate Analytics) searches were combined.Comments, reports, technical notes, letters, and videos were also removed.Initially, two authors (FJ and HV) independently screened all studies for relevance and eligibility based on the title and abstract.The authors resolved disagreements through discussion and consensus.Articles remaining after filtering based on the abstract were considered for full-text evaluation.Two authors (FJ and HV) completed several independent rounds of verification before deciding which studies to consider for review.Finally, a secondary search was conducted by two

Information synthesis
The included references were divided into different groups according to the simulation method of the aneurysm and neighboring anatomies as seen in Table 2.
As no quantitative studies were found, the following aspects of each paper are descriptively highlighted: 1 Summary of the study method: This included the number and type of simulated intracranial aneurysms, the number and background of participants if mentioned, performed validity assessment, reported learning/simulation outcomes, and reported disadvantages.Regarding validity, it was noted whether the training method was tested for face validity, content validity, construct validity, feasibility, predictive validity, or concurrent validity.Face validity is the extent to which the simulation resembles real life, content validity refers to the extent to which the simulation is complete and accurate, and construct validity is the extent to which performance on the simulator discriminates between novices and experts [11].Feasibility refers to the measurement of whether the simulation is possible, predictive validity involves checking how well the performance on a simulator predicts future performance, and concurrent validity is the extent to which performance on a simulator correlates with best practices.
The simulation approach, study design, and validation method of each reference were assessed; the findings are summarized in Tables 3, 4, and 5.

VR
Different setups were used for the VR simulations.Some studies used a surgical rehearsal platform such as Surgical Theater (Cleveland, OH, USA) [19,20,24,28]         Several studies created a custom-designed VR setup [21,22].

3D static simulators
Pathology is replicated using 3D printers.Apart from the aneurysm and the neighboring vessels, some studies also included the skull and/or brain.The studies focused on creating hollow aneurysms [31] or a total hollow vasculature [32][33][34] in the model by using elastic 3D-printed material [30,35].

3D dynamic simulators
All dynamic simulators replicated the aneurysm and vasculature with hollow elastic vessels, allowing for blood flow simulation.Leal

Method of assessment
The assessment method for the simulation approach varied across articles.Most articles included a subjective, descriptive questionnaire to assess the face validity, content validity, and/or feasibility [9, 13-19, 21-23, 26, 30-34, 36].Some articles on VR simulation compared a control group with a VR group; the control group used only conventional presurgical planning tools, while the VR group used a VR simulator to perform presurgical planning [19,20,24,28].
A comparison of aneurysm complexity, surgical time, and surgical outcomes was made [19,20], or surgical videos were recorded, which allowed for the assessment of the clipping itself by an experienced neurosurgeon [24,28].
In other studies that conducted a review, a senior neurosurgeon assessed clipping performance based on neurosurgical videos [13] or microscope videos of the simulation [36] or by watching the residents' performance [33].Belykh et al. (2017) [15,17] used an objective tool to assess the construct validity, namely the Objective Structured Assessment of Aneurysm Clipping Skills (OSAACS) tool, which measures specific operative nuances of aneurysm clipping surgery for which an assessor determines the performance of residents.Another assessment method used by some references involved a comparison of the virtual clipping and the surgical clipping based on the clip used, clip position, and the number of clips [22,23,25,27,29,35,37].

Quality assessment
The articles varied in quality, which was reflected in their assessment methods.Some references only included a descriptive questionnaire for assessment, which is a subjective assessment leading to lower quality outcomes.Most references did not mention potential bias in their study.Belykh et al. (2017) [15] noted the presence of some bias as the raters were not blinded to the participants.Kockro et al. ( 2016) [25] reported selection bias as the complex aneurysm cases were mostly handled by experienced neurosurgeons.Marinho et al. (2014) [27] reported analysis bias due to the subjectivity of the limited number of participants.Some studies used blinding to decrease the risk of bias.In the studies by Oliveira et al. (2018) [13] and Chugh et al.
(2017) [24], the reviewer was blinded as to whether conventional planning or simulation was used before the surgery.In the study by Joseph et al. (2020) [36], the microscope videos were anonymized so that the reviewer was blinded to the participant's identity.The study by Bambakidis et al. (2013) [28] was double blinded.Additional sources of bias are observed by the authors.Some studies included only a limited number of participants [23,24,27,31,33,37], leading to bias regarding the level of experience.Other studies simulated only a low number of aneurysms [21,26,28,30,36,37], resulting in a bias toward the complexity of the aneurysm case.It is not sure if the studies would have the same outcomes with surgeons of different experience and with other aneurysm complexity cases.

Summary
The creation of good training methods for the clipping of cerebral aneurysms is crucial for the proper education of trainees.Different methods have been investigated to recreate aneurysms based on ex vivo specimens, including the use of chicken wings, pigs, cadaveric brains, 3D-printed aneurysms implanted in cadavers, and cadaveric cow craniums [38][39][40][41][42]. Aboud et al. (2015) [14] and Oliveira et al. ( 2018) [13] reported using human cadavers with and without blood flow and dynamic behavior.These methods provided excellent haptic feedback and allowed the residents to train in multiple surgical steps, but the studies did not report definitive outcomes.Some cadavers with pathology were collected over an extended period and were difficult to perfuse.This type of setup is limited because cadavers are resource intensive and require special refrigeration, making them unsuitable for extended resident training programs.In addition, the studies did not report on the repeatability of procedures or the feasibility of patient-specific learning outcomes.The studies by Belykh et al. and others that used human placentas and chicken wings [15][16][17][18] included dissection and perfusion to replicate dynamic behavior but failed to replicate the surgical workflow during and after the clipping procedure.They included a structured performance assessment by an expert, which proved that the trainees had learned.Oliveira et al. (2018) [13] showed that residents training on placenta models performed better than residents training on cadavers or only having surgical videos available.However, without human-specific anatomy and pathological geometry in ex vivo models, these studies could not compare trainee performance to the clipping approach.In addition, preparing such ex vivo models is time-consuming [14], so they are unsuitable for presurgical planning and training.VR systems do not have preservation problems, and they have been highly investigated for planning [43][44][45], focusing on force feedback, distance visualization, or a craniotomy approach [46][47][48].Participants in VR studies could clip aneurysms multiple times, but training for the surgical experience was unrealistic although the simulation created an immersive environment.These simulators lack realistic haptic feedback, tactility, pulsatile flow, and real-life simulation [19,23,[27][28][29].Some VR studies implemented haptic feedback through haptic controllers for both hands, but participants scored haptic reality low [21,22,24,26].All VR studies could simulate aneurysm clipping, but none could include all other steps involved in a complete surgical workflow (Table 4).Availability, cost, and access to VR simulators created barriers to their use [19,23].The VR studies that compared a control group showed decreased surgical time when training with a VR simulator [20,24,28].
More practitioners are using 3D-printed models to understand clipping procedures in their education and training because technology has progressed rapidly and patientspecific approaches are made possible by producing models from patient images [49][50][51][52][53][54].Hollow aneurysm models have been created, which allow for the performance of clipping [55,56], but they lacked realistic haptic touch.The models did not introduce blood flow or surgical complexities.Because of these missing components, only a limited part of the surgical workflow could be simulated.Although static training models have these major limitations, an advantage is that 3D models can be reused for multiple training sessions if no craniotomy is performed.
The absence of the dynamic behavior of blood flow is a big disadvantage as it does not allow for the confirmation of a successful clipping.The clip should occlude the aneurysm completely at neck level, and parent arteries should not be compromised by the clip [57].This can be confirmed when blood flow is available in the training model.Adding a contrast marker in the blood can facilitate the validation of correct clipping under a microscope.By using pulsatile flow, touching the dome and arteries can already give an idea of whether the clipping was successful, and the use of Doppler angiography can provide further confirmation.
This leads to dynamic 3D simulations being more realistic as they allow simulation of the blood flow [58,59].Joseph et al. (2020) [36] and Liu et al. (2017) [9] added pulsatile blood flow and realistic haptic feedback to their simulation models, allowing for a realistic simulation of the aneurysm approach and clipping as was reported by the participants of these studies.These dynamic models can be reused for multiple training sessions, making them a cost-effective option.Besides these advantages, the dynamic models are not realistic for the whole neurosurgical workflow as they also lack skin, meninges, and important microanatomical features such as cranial nerves, smaller arteries, and veins.Aneurysm clipping surgery is not only about the clipping itself but also involves the planning approach, strategy, and direction to safely reach the pathology without sacrificing major microanatomical structures.Therefore, the dynamic simulators are not sufficient to simulate important microsurgical steps.Future research should focus on simulations that include all anatomical and microanatomical structures to have a more valid representation of the complete surgery.Different technological methods could be combined to address some limitations, for example, creating a mixed environment using augmented reality technology and incorporating dynamic flow on VR system.

Limitations
The heterogeneity of the reviewed studies made it difficult to perform direct comparisons.This is related to the variance in training and assessment methodology worldwide.A standard methodology scale could be used to objectively determine clipping scores.An effort for this was already performed by creating the Objective Structured Assessment of Aneurysm Clipping Skills (OSAACS) [15] and Skills Assessment in Microsurgery for Brain Aneurysms (SAMBA) [60] scoring systems.However, both still have limitations in assessing the complete aspects of aneurysm surgery.SAMBA focuses only on the aneurysm clipping aspects, while OSAACS focuses more on other intrasurgical steps and less on the clipping itself.Therefore, a combination of these two scoring systems should be developed.For future research, a consensus on assessment methods should be reached to compare the different studies.Implementing hands-on training is limited by different factors; the availability of the simulators, the supply chain, the replacement of head micro-anatomy when ruptured, the training site's economic situation, and mentorship availability all play an important role in establishing the training environment [61][62][63].No clear validation method is available for the training methods; therefore it is impractical for studies to prove their effectiveness and to make a valid comparison with other training methods.In the future, similar studies should focus on comparing different simulation methods and comparing them to a control group without simulation to be able to show effectiveness.

Future directions
The overall significance of this systematic review lies in developing the latest knowledge on patient-specific realistic training approaches to be able to educate neurosurgeons and residents and help researchers and medical device manufacturers in designing, developing, and implementing simple and innovative systems in treatment practices.The key perspectives represented in this review could help stakeholders like surgical staff, hospital administration, medical universities, and training institutes to understand the evolution of training and the need to adapt to future directions during the development of policies and guidelines for the aneurysm training and treatment process.

Fig. 1
Fig. 1 PRISMA flow chart of the selection process for the included studies

17 [ 27 ]
created a virtual operative environment with the opensource software Blender.Alaraj et al. (2015) [26] developed a VR simulator using the ImmersiveTouch (Chicago, IL, USA) platform.Shono et al. (2018) [23] made use of the integrated development environment of unity to create the simulation.
et al. (2019) [37] injected a physiological solution into the aneurysm to represent blood flow.Joseph et al. (2020) [36] used a custom-designed pulsatile pump, and Liu et al. (2017) [9] used a blood flow driver to simulate pulsatile blood flow.

Table 1
Search strategy (MEDLINE)

Table 2
Different types of training approaches

Table 3
Summary of the study methods

Table 5
Applicability of simulation