Simulation speeds up adaptation to the 3D exoscope: a randomized controlled trial Authors: Adam Yousfi, Ville Vasankari, Anni Pohjola, Emil Lintunen, Elias Roos, Martin Lehecka Published: 2026 ABSTRACT IntroductionExoscopes are alternatives to operative microscopes in intraoperative visualization. Transitioning from microscope to exoscope demands new skills. Simulators are effective training tools in multiple fields, including surgery, yet no virtual simulators for exoscope training exist.Research questionWe developed a virtual exoscope simulator and evaluated whether short-term simulator training improves exoscope control and microsurgical task performance in novices.Material and methodsForty medical students without prior surgical experience were randomized to a control group or 45 min of simulator training focused on exoscope control. Both groups then performed a standardized microsurgical test task using a real 3D exoscope. The task was a custom-made physical model requiring participants to transit between concealed metal discs (diameter 0.5 cm) at different depths and visual angles while passing a microneedle with 8-0 thread through them, designed to simultaneously assess exoscope camera control and manual microsurgical performance. All participants completed the task three times in a single session.ResultsThe simulator group demonstrated superior exoscope control by transitioning between the discs faster (median 4min 26s vs. 5min 17s, p = .009) and entering discs faster (median 4min 8s vs. 8min 1s, p = .0006) than the control group. Both groups improved across trials. The standardized effect size was moderate (Cohen's d = 0.56). Groups did not differ significantly in number of exoscope movements or needle drop frequency.Discussion and conclusionBrief simulator training enhanced adaptation to exoscope controls and manual task performance in novices. Further studies involving neurosurgeons are needed to determine effect on users who possess established microsurgical skills. Introduction Digital 3D exoscopes are novel alternatives to microscopes in intraoperative visualization (Langer et al., 2020; Montemurro et al., 2022; Moisi et al., 2017; Muhammad et al., 2019; Zhang et al., 2022). Exoscopes have been successfully applied in complex neurosurgical procedures, resulting in similar outcomes as with traditional operating microscopes (Auricchio et al., 2024; Calvanese et al., 2024; Rossmann et al., 2023; Veldeman et al., 2023). Although the surgical techniques used in exoscopic surgery are identical to those used with a microscope, switching from microscope oculars to a three-dimensional exoscope display and the adaptation to exoscope controls, e.g. the footswitch that controls the robotic arm, presents ergonomic and perceptual challenges and requires training. Currently, no virtual simulators for exoscope training are available. Exoscope training takes place mainly in the operating room during live cases, which raises ethical concerns. Poor early experiences may negatively affect surgical performance and also result in reluctance to continue using the device. Cadaver and animal laboratory courses are costly and logistically demanding. There are currently no safe, unlimited practice options allowing automated skill assessment for exoscopes. Previous studies have demonstrated that videogaming experience in general correlates with better adaptability to the surgical exoscope (Gupta et al., 2021; Lynch et al., 2010; Yousfi et al., 2025a), suggesting that a dedicated exoscope simulator could provide similar benefits. We developed an exoscope simulator to train effective control of surgical exoscopes. The aim of this study was to evaluate whether short-term simulator training improves adaptation to a real 3D exoscope compared with no simulator training, assessed using a standardized microsurgical test task performed under an exoscope. We hypothesize that: (1) simulator training leads to superior adaptation to exoscope compared with no simulator training; (2) overall manual skills are similar between groups; and (3) repetition of test task will lead to learning effect unrelated to the initial training. Materials and methods Forty medical students without prior surgical experience were randomly assigned to receive: (a) 45 min of simulator training on exoscope use (simulator group) or (b) no simulator training (control group). The simulator was custom built to facilitate the adaptation to exoscope controls. Following simulator training, both groups performed a standardized microsurgical test task using a real footswitch-controlled 3D exoscope. The task consisted of passing a microneedle with 8-0 thread in a defined order through six metal minidiscs mounted on an asymmetric surface of a custom-built 3D printed model. All participants from both groups performed the same microsurgical test task three times in a row during a single test session. Participants We recruited 40 medical students (27 women, 13 men) in the 1st-3rd year of their studies. None had prior experience in microsurgery or working with exoscopes. All participants reported normal three-dimensional depth perception. Data collection took place between June and August 2025. Written informed consent was obtained from all participants. The study was conducted in accordance with the Declaration of Helsinki. Study design Participants were assigned either to the simulator or control group in a 1:1 ratio using a computer-generated randomization sequence created in SAS prior to enrollment. Allocation was implemented sequentially according to order of enrollment, resulting in assignment to either simulator training before the test task (n = 20; 13 women, 7 men) or no simulator training (n = 20; 14 women, 6 men). Both groups were provided with a short instructional video on exoscope use. The simulator group practiced at the Helsinki University Hospital Microneurosurgery Laboratory with the ExoscopeTwin simulator (version 1.0) for 45 min, after which they started the test task. The test task, performed with the 3D exoscope (Aesculap AEOS®, BBraun), was preceded for both groups by a brief, on-location tutorial on exoscope use, and 15 min of training. The tutorial was conducted by one researcher, AY, who was not blinded to group allocation due to involvement in the simulator training. The researcher explained the exoscope controls and demonstrated the location of all target discs followed by a brief lesson on microsurgical needle-handling. During the training, the trainees visualized the discs in correct order and passed a microneedle once through a single disc, prior to starting the test task. All test tasks were video recorded and stored in an external hard drive for subsequent analysis. A schematic drawing of the study execution is presented in Fig. 1.Fig. 1CONSORT flow diagram showing the enrollment, allocation, follow-up, and analysis of participants. All 40 enrolled participants completed the study protocol and were included in the final analysis. CONSORT flow diagram showing the enrollment, allocation, follow-up, and analysis of participants. All 40 enrolled participants completed the study protocol and were included in the final analysis. Simulator training (simulator group) In the experimental setup, the simulator group trained using a custom-developed exoscope simulator (ExoscopeTwin, version 1.0), designed and built by AY, ER, and EL in consultation with ML. The main aim of the simulator is to teach the different controls and functions needed in fluent use of the exoscopic camera. It is not trying to teach microsurgical skills as such. The simulator consists of three components: (1) custom software, (2) a custom replica of an exoscope footswitch, and (3) a standard computer mouse (Video 1). Supplementary video related to this article can be found at doi:10.1016/j.bas.2026.106126 The following is the supplementary data related to this article:Multimedia component 2Representative recording of author AY demonstrating simulator use. Multimedia component 2 Representative recording of author AY demonstrating simulator use. The participants viewed the simulation on a large 32″ screen positioned directly in front of them, replicating the experience of using a real exoscope. The simulator interface closely resembled the visual output of an exoscope camera (Fig. 2a, Fig. 2b, Fig. 2c, Video 1). The simulator footswitch was used for planar translations, rotational movement, and adjustment of focus and magnification in a manner analogous to the real exoscope footswitch (Fig. 3). Although the simulator allows user-specific programming of footswitch settings, in this experimental setup the pedal layout was standardized for all participants to replicate the commonly used clinical setup.Fig. 2aThe simulator runs on a personal computer and is displayed on an external monitor via HDMI. A virtual 3D training model is visible on the screen. A standard computer mouse is used to manipulate a virtual microinstrument, which trainees use to drag numbered targets on the surface of the simulated 3D model.Fig. 2bA virtual 3D training model created to practice exoscope maneuvering skills, zoom, and manual focusing ability.Fig. 2cA photogrammetry 3D brain scan model used for training in the simulator environment.Fig. 33D-printed functional replica exoscope footswitch. The simulator runs on a personal computer and is displayed on an external monitor via HDMI. A virtual 3D training model is visible on the screen. A standard computer mouse is used to manipulate a virtual microinstrument, which trainees use to drag numbered targets on the surface of the simulated 3D model. A virtual 3D training model created to practice exoscope maneuvering skills, zoom, and manual focusing ability. A photogrammetry 3D brain scan model used for training in the simulator environment. 3D-printed functional replica exoscope footswitch. The training software contained multiple neurosurgical models, ranging from the abstract to the realistic. The trainees progressed from the abstract models to the more realistic ones during the training session. The simulation was gamified to increase trainee focus. The models were specifically designed to gradually teach different skills: initial models focused separately on planar movement, camera tilting, focus adjustment, or magnification optimization, while the advanced models combined all of these tasks into a single, integrated challenge. Visual cues had to be followed by manual execution using a computer mouse. Test task (both groups) The test task was designed to assess active exoscope use, including complex camera movement and image adjustments, as well as fine micromotor skills. All participants performed the same standardized test task three times during a single session, with a 5-min break between the repetitions. We used a custom-developed 3D-printed model for the test task (Fig. 4a, Fig. 4b, Fig. 4c, Fig. 4d, Fig. 4e). It was specifically designed to require active and frequent exoscope camera manipulation (Video 2). The model features a narrow orifice resembling a keyhole surgical approach and contains six metal minidiscs (0.5 cm diameter) positioned at different depths and orientations. Three platforms at different depths are located within the model, each containing a concealed minidisc. Three discs are also located on the outer surface of the model. Each disc contains four symmetrically positioned slits through which a microneedle may pass. The visualization of the different discs requires heavy tilting of the exoscope camera.Fig. 4aThe model used in the test tasks as viewed by a trainee on the exoscope display.Fig. 4bA 3D representation of the model.Fig. 4cOn-screen areas used during video analysis.Fig. 4dNeedle first touches the target disc. In this case, zoom would be 1, target centrality 1 and disc visibility 1/3.Fig. 4eNeedle exits the target disc. In this case, zoom would be 1, target centrality 1 and disc visibility 1/3. The camera angle remains identical to Fig. 4d. The model used in the test tasks as viewed by a trainee on the exoscope display. A 3D representation of the model. On-screen areas used during video analysis. Needle first touches the target disc. In this case, zoom would be 1, target centrality 1 and disc visibility 1/3. Needle exits the target disc. In this case, zoom would be 1, target centrality 1 and disc visibility 1/3. The camera angle remains identical to Fig. 4d. Supplementary video related to this article can be found at doi:10.1016/j.bas.2026.106126 The following is the supplementary data related to this article:Multimedia component 3Part of the test task. At the start of the video, the trainee has already passed the needle through disc 1 and is now visualizing a predefined point on the concealed inner wall of the model (treated as disc 2 only for time and camera movement parameters). The video ends with the trainee passing the needle through disc 3. The task requires the trainee to still pass the needle sequentially through discs 4–7, but this part is not depicted in the video. Each target disc must be visualized from the outside. The needle must pass entirely through each disc, but the needle may be withdrawn immediately after full entry. Multimedia component 3 Part of the test task. At the start of the video, the trainee has already passed the needle through disc 1 and is now visualizing a predefined point on the concealed inner wall of the model (treated as disc 2 only for time and camera movement parameters). The video ends with the trainee passing the needle through disc 3. The task requires the trainee to still pass the needle sequentially through discs 4–7, but this part is not depicted in the video. Each target disc must be visualized from the outside. The needle must pass entirely through each disc, but the needle may be withdrawn immediately after full entry. The task objective was to pass a needle attached to an 8-0 monofilament thread through all six discs in numerical order using two jeweller's forceps. During the test task, the participants were allowed to freely adjust the zoom and focus of the camera view, as well as reposition the exoscope camera whenever needed using the footswitch. Exoscope handling was assessed based on the speed and accuracy of transitions between discs, while manual skill performance was evaluated by the ability to rapidly pass the needle through each disc without dropping the needle or moving the forceps out of the surgical field. Data analysis Automatically recorded simulator performance data were analysed using the parameters listed in Table 1. Video recordings of the exoscope test tasks were analysed manually by an independent researcher, AS, blinded to group allocation and participant identities. Video analysis focused on two main areas: a) exoscope handling and b) fine motor skills. Time was measured from the start of the task (first exoscope movement) until completion (needle passes through the last disc). Additional timestamps were recorded for transitions between the discs. All parameters used for the analysis of the test task are presented in Table 2. The primary outcome was total task duration, defined as the time from the first exoscope movement to completion of the task. Secondary outcomes included transition times between discs, time taken to pass the needle through discs, exoscope camera movements, and task errors (needle drops and instruments out of the field).Table 1Summary of quantitative variables describing the use of different exoscope functions during simulator training in the simulator group.Table 1Completed TasksNumber of tasks completed in the simulation in 45 minMovementTotal number of camera movementsMovement ModeTotal number of switches between movement-in-plane and viewing-angle tilt modesZoomNumber of times zoom was usedFocusNumber of times manual focus was usedAutofocusNumber of times autofocus was usedNeedle ResetNumber of times the virtual needle was automatically recentredTable 2Exoscope test task variables. See Fig. 4a, Fig. 4b, Fig. 4c, Fig. 4d, Fig. 4e for visual clarification on test task structure and grading.Table 2Total TimeDuration of the taskTransition TimeTime spent transitioning between the discsDisc Entry TimeTime taken to pass the needle through discsCamera Movement, TransitionNumber of exoscope camera movements (movements in plane and tilting movements) when transitioning to discs.Camera Movement, Disc EntryNumber of exoscope camera movements when passing the needle through discs.Disc VisibilityVariable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc (1/3: <33 % of the disc visible, 2/3: 33-66 % visible, and 1 > 66 % visible). Variable value is the sum of all measurements.ZoomSize of unobstructed disc on screen (estimated by the number of squares in the overlaid grid). Variable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc. Variable value is the sum of all measurements.Target CentralityHow well is the camera centred on the disc? Which area of the overlaid grid, 1-3, contains the disc? Variable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc. Variable value is the sum of all measurements.Needle DroppedNumber of times that the needle was dropped.Instruments out of FieldNumber of times that both instruments were out of the surgical area. Table 1Summary of quantitative variables describing the use of different exoscope functions during simulator training in the simulator group.Table 1Completed TasksNumber of tasks completed in the simulation in 45 minMovementTotal number of camera movementsMovement ModeTotal number of switches between movement-in-plane and viewing-angle tilt modesZoomNumber of times zoom was usedFocusNumber of times manual focus was usedAutofocusNumber of times autofocus was usedNeedle ResetNumber of times the virtual needle was automatically recentred Summary of quantitative variables describing the use of different exoscope functions during simulator training in the simulator group. Table 2Exoscope test task variables. See Fig. 4a, Fig. 4b, Fig. 4c, Fig. 4d, Fig. 4e for visual clarification on test task structure and grading.Table 2Total TimeDuration of the taskTransition TimeTime spent transitioning between the discsDisc Entry TimeTime taken to pass the needle through discsCamera Movement, TransitionNumber of exoscope camera movements (movements in plane and tilting movements) when transitioning to discs.Camera Movement, Disc EntryNumber of exoscope camera movements when passing the needle through discs.Disc VisibilityVariable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc (1/3: <33 % of the disc visible, 2/3: 33-66 % visible, and 1 > 66 % visible). Variable value is the sum of all measurements.ZoomSize of unobstructed disc on screen (estimated by the number of squares in the overlaid grid). Variable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc. Variable value is the sum of all measurements.Target CentralityHow well is the camera centred on the disc? Which area of the overlaid grid, 1-3, contains the disc? Variable analysed twice for each disc, once when the needle first contacts the disc, and once when it exits the disc. Variable value is the sum of all measurements.Needle DroppedNumber of times that the needle was dropped.Instruments out of FieldNumber of times that both instruments were out of the surgical area. Exoscope test task variables. See Fig. 4a, Fig. 4b, Fig. 4c, Fig. 4d, Fig. 4e for visual clarification on test task structure and grading. Statistical analysis Statistical analyses were conducted using SAS software, version 8.3, with the PROC MIXED procedure for mixed-model analysis and Spearman correlation analysis. Quantitative variables were treated as continuous. Following the primary analyses, additional exploratory analyses were performed to support interpretation of the findings. These included calculation of standardized effect sizes (Cohen's d), sensitivity power calculations based on observed means and variances, and a nonparametric bootstrap resampling approach (1000 iterations) applied to the primary outcome (Total Time) to assess the stability of the estimated between-group difference. All tests were two-sided, and a p-value ≤0.05 was considered to indicate statistical significance. Due to the skewed distribution of outcome variables, descriptive statistics are presented as median and interquartile range. PROC MIXED was used for inferential analysis as it accommodates non-normal residuals in repeated-measures designs. Secondary outcomes, including transition times, camera movements, disc visualization parameters, and task errors, were exploratory in nature and were not corrected for multiple comparisons. Findings should be interpreted with appropriate caution. Results Completion time and task phases The simulator group was faster than the control group in the exoscope test task, while both groups improved their times during the multiple test runs (Fig. 5 and 6a). Median task duration improved from 11 min 23 s (IQR 10 min 18 s) to 8 min 5 s (IQR 5 min 29 s) in the simulator group, and from 16 min 59 s (IQR 15 min 14 s) to 10 min 24 s (IQR 6 min 37 s) in the control group. There was a significant difference between the groups in favour of the simulator group (p = .0003). Additionally, both groups showed significant improvement across the trials (p < .0001), while both groups improved similarly (p = .1).Fig. 5Synthesis of variable averages across all trials. See Table 2 for details on the variables. For the purposes of this figure, Camera Movement Total and Zoom—values have been divided by five and Target Centrality with three to fit the data into the graph. Time is reported in minutes.Fig. 6aSummary statistics for time measurements (in seconds) across trials in the simulator and control groups. Synthesis of variable averages across all trials. See Table 2 for details on the variables. For the purposes of this figure, Camera Movement Total and Zoom—values have been divided by five and Target Centrality with three to fit the data into the graph. Time is reported in minutes. Summary statistics for time measurements (in seconds) across trials in the simulator and control groups. Given the modest sample size (n = 20 per group) and limited sensitivity power for Total Time (≈0.41), a nonparametric bootstrap analysis (1000 resamples) was performed to assess the robustness of the observed effect. The mean bootstrap-estimated between-group difference in change from Trial 1 to Trial 3 was 282 s, with an interquartile range of 193–366 s, and 98.9 % of bootstrap iterations favored the intervention. The standardized effect size was moderate (Cohen's d = 0.56). Although smaller effects may have gone undetected and generalizability is limited, the consistency of the bootstrap estimates and the magnitude of the effect support a meaningful benefit of the intervention. Larger studies are warranted to confirm these findings and refine effect size estimates. When analysing specific task phases, the simulator group demonstrated better exoscope control by transitioning between discs faster than the control group (p = .009; Figs. 5 and 6a). Furthermore, the simulator group demonstrated superior fine motor skills, passing the needle through the discs faster (p = .0006; Figs. 5 and 6a). Both groups showed significant improvement in these sub-skills across trials (p < .0001), with similar rates of improvement between groups (p = .18 for transitions between discs and p = .19 for passing needle through discs, respectively). Exoscope camera movements Camera tilting was used by both the simulator and control groups equally often (p = .48; Figs. 5 and 6b). Both groups showed a significant reduction in the number of movements across the trials (p < .0001), and the rate of improvement was similar in both groups (p = .997).Fig. 6bSummary statistics for number of exoscope camera movements across trials in the simulator and control groups. Summary statistics for number of exoscope camera movements across trials in the simulator and control groups. Similar results were obtained when examining separately the number of exoscope movements required to pass the needle through the discs and transit between them. In both cases, movement became more efficient in both groups following the multiple test runs. No statistically significant differences between the groups or the trials were observed in zooming in and out, disc visualization and centring of the target disc (Figs. 5 and 6c).Fig. 6cSummary statistics for image quality metrics across trials in the simulator and control groups. Summary statistics for image quality metrics across trials in the simulator and control groups. Although camera movement frequency did not differ significantly between groups, the simulator group demonstrated shorter inter-disc transition times and shorter time to pass the needle through discs compared to controls, while both groups achieved similar visualization of the discs. Instrument control and fine motor errors Participants in either group rarely lost the control of their needle. The simulator group dropped the needle less often than the controls (p = .035; Figs. 5 and 6d). The number of needle drops did not change between consecutive tasks (p = .22).Fig. 6dSummary statistics for instrument control across trials in the simulator and control groups. Summary statistics for instrument control across trials in the simulator and control groups. The control group took both instruments out of the optimal viewing field more often than the simulator group (p = .018; Figs. 5 and 6d). However, the frequency of these events did not change between consecutive trials (p = .31). Simulator vs. test task performance Overall, simulator performance metrics did not strongly predict success in the subsequent exoscope test task. However, certain parameters showed indicative trends. A high frequency of transitions between the movement modes (in-plane movement vs. tilting around the target) during the simulation showed a weak positive correlation with disc visibility (ρ = 0.25, p = .05) and a weak negative correlation with the target centrality (ρ = −0.28, p = .03) in the exoscope test task (note that smaller values in the target centrality parameter indicate that the discs are located more centrally on the exoscope screen). Additionally, a high frequency of manual focus use in the simulation was weakly correlated with longer transition times between discs (ρ = 0.25, p = .05), whereas frequent autofocus use showed a weak negative correlation with transition time (ρ = −0.26, p = .05). Discussion Supporting our first hypothesis, the simulator-trained participants had superior overall performance in all three trials of the exoscope test task. They demonstrated superior control of the exoscope by transiting between the discs faster while gaining similar visuals on the targets compared to the controls. Contrary to our second hypothesis, manual dexterity differed between the groups to the benefit of the simulator group, as they passed the needle through the discs faster. This may reflect enhanced perception of manual actions on the exoscope screen or increased confidence rather than direct manual skill acquisition. Supporting our third hypothesis, the novice users from both groups adapted quickly to the surgical exoscope and improved in both controlling the exoscope and fine micromotor skills during the test runs. Two findings from the present study warrant closer interpretation. First, despite transitioning between the discs significantly faster, the simulator group did not perform fewer total camera movements than the controls, and both groups achieved comparable disc visualization quality throughout. This dissociation suggests that simulator training did not reduce the quantity of camera adjustments but rather improved their quality. Each movement was likely more purposeful and spatially accurate, reducing the time spent searching for the target disc and minimizing corrective micro-adjustments after each repositioning. This has direct clinical relevance given that cumulative camera repositioning time contributes meaningfully to total operative time in live surgery. Second, neither needle drops nor having instruments-out-of-field improved significantly across trials in either group, despite substantial reductions in overall task duration. This apparent paradox likely reflects a combination of low baseline error frequencies limiting further detectable improvement, and a speed-accuracy tradeoff whereby a faster pace across trials offset the precision gains otherwise expected with practice. These observations highlight that time-based and error-based outcomes capture qualitatively distinct dimensions of microsurgical performance, and both should be reported in surgical simulation research. Short exoscope courses have been found efficient previously too (Giammalva et al., 2023; Vasankari et al., 2024). In a laboratory study of ten neurosurgeons, six short training sessions with a microvascular anastomosis model led to major improvements in exoscope handling and manual dexterity, especially among the less-experienced surgeons (Vasankari et al., 2024). Another study demonstrated that 18 neurosurgeons improved their speed and accuracy within just three repetitions in a simulated exoscope-assisted neurosurgical task independent of prior microscope experience (Giammalva et al., 2023). Similarly, in a retrospective series of 24 exoscope-assisted spinal procedures, surgeons achieved proficiency after only a few cases, with operative times significantly decreasing after the initial learning phase (Milani et al., 2024). These findings complement our results and highlight the rapid learning curve for the use of modern 3D exoscopic systems. Virtual simulators in surgery Simulation-based training has become an essential component of multiple technical fields, including aviation, engineering, and industrial safety training. In medicine, simulators are increasingly being integrated into training across specialties, although medical simulation training programs remain often institution specific. Surgical simulators can be classified into two main categories: 1) mechanical simulators coupled with a) synthetic models or b) cadaveric or animal models, and 2) virtual simulators. The training of microsurgery has traditionally been based on a master apprentice setting while operating on real patients. Nowadays, essential motor skills, such as microinstrument handling and suturing, can be practiced independently in a microsurgical laboratory (Atkins et al., 2005; Trignano et al., 2017; Vasankari et al., 2024; Choque-Velasquez et al., 2018; Yousfi et al., 2025b; Schildt et al., 2026). Most recently, simulation-based approaches have been designed to enhance neurosurgical education (Stredney et al., 2013; Wang et al., 2006; Lemole et al., 2007; Delorme et al., 2012). For example, virtual simulators are available for the practicing of drilling, tumor debulking, and tissue cutting and retraction. For the exoscope, several synthetic (e.g. SurgTrain®, VIRVELI) and animal models have been developed (Vasankari et al., 2024; Yousfi et al., 2025a). However, their use requires the physical exoscope device itself, along with dedicated space and equipment setup. The ExoscopeTwin simulator developed for this study is the only virtual exoscope simulator that we are aware of. A key strength of the simulator is its compact design, which prioritizes familiarization with device handling over all-encompassing procedural realism. This device-oriented approach effectively supported early-stage skill acquisition. In other surgical fields, the use of standardized virtual simulators in training is well established. Multiple virtual simulators have been developed for the training of endoscopic surgery during the last decades (van der Wiel et al., 2016; Toale et al., 2023; Yang et al., 2018). Simulator-based training is especially recommended for basic flexible endoscopy (esophagogastroduodenoscopy, sigmoidoscopy and colonoscopy), endoscopic retrograde cholangiopancreatography and laparoscopy (Ekkelenkamp et al., 2016). Virtual simulators have also been designed for and found beneficial in robotic surgery, for instance, virtual simulation facilitates adaptation to the daVinci surgical system, but the costs of the simulation equipment are often considered too expensive compared to its benefits (Tanaka et al., 2016; Foell et al., 2013; Whittaker et al., 2016). From the virtual surgical simulators, EyeSi®, developed for intraocular surgery, resembles our simulator the most as it employs an almost similar footswitch mechanism to the one used to control the AEOS surgical exoscope (Liu et al., 2024; Jaud et al., 2021; Carr et al., 2024). The EyeSi has been found to improve technical skills, procedural proficiency, and theoretical knowledge in cataract, vitreoretinal, and fundus examination training. Strengths As far as we know, this is the first study on a virtual exoscope simulator. Second, our standardized and easily replicable 3D printed test task allowed objective assessment of both microsurgical dexterity and exoscope handling simultaneously, addressing a training dimension that is frequently neglected in many existing microsurgical models. Third, the simulator and test task were reviewed by multiple experienced neurosurgeons, who found them useful and appropriately challenging, especially when time pressure was applied. Fourth, high-quality video recordings enabled detailed analysis of the exoscope test task. This was complemented by the simulator's automatic recording of various parameters, providing valuable data on its use. Fifth, both groups received 15 min of guided training with the real exoscope before commencing the test task. Despite this, the simulator group significantly outperformed the control group. Sixth, the use of surgical novices ensured that the results reflect pure learning effects rather than prior experience. Seventh, although participants were not stratified based on sex, prior video gaming, handicraft, or 3D modeling experience, these characteristics were roughly balanced between the simulator and control groups. Limitations This study has several limitations. First, we did not assess the long-term accumulation of exoscope skills following simulator training, which may affect the generalizability of the results over time. Second, all tasks were conducted in a controlled laboratory environment with novice participants. All participants were medical students with no previous experience in surgery, microsurgery, or exoscope use. Consequently, the observed effects represent the learning process of complete novices and cannot be directly extrapolated to neurosurgical residents or practicing neurosurgeons. We deliberately selected novices because experienced neurosurgeons already possess highly developed microscope-handling skills, and part of the challenge of adopting a new visualization platform may involve unlearning established habits associated with conventional microscope use. Using complete novices allowed us to minimize this potential source of bias and isolate the process of acquiring exoscope-control skills. However, this choice introduced a different limitation, as the participants lacked any prior microsurgical experience. Our primary objective was not to assess microsurgical proficiency but rather the acquisition of skills required to operate a novel visualization device. Based on our empirical observations to date, extensive prior microscope experience does not appear to substantially accelerate the learning of exoscope use, although this question warrants further investigation. The learning process itself was accounted for in the results, and the intervention group outperformed the control group in all trials: both groups improved similarly, but the intervention group consistently performed better. Nevertheless, the findings may not reflect performance in actual surgical cases among residents and neurosurgeons. Third, the simulator could be further refined to better mimic real surgical conditions, for example by incorporating more realistic tissue interactions. Fourth, the simulator setup was based on a 2D screen. The results might have differed if participants had used a simulator version compatible with VR glasses, enabling the viewing of exoscope camera output from a virtual 3D screen. Fifth, sample size, while adequate for initial analysis, could be expanded in future studies to enhance statistical power and robustness. Sixth, although the videos were graded by a blinded researcher, the researcher who trained students on exoscope and simulator use was not blinded. Conclusions The virtual exoscope simulator appears to facilitate adaptation to the surgical exoscope among novice users by improving their ability to control the exoscope camera and perform manual tasks when viewing their actions from a screen. Further studies involving neurosurgeons could determine whether the simulator improves device-specific adaptation in users who already possess established microsurgical skills. Ethical approval The manuscript does not contain clinical studies or patient data, which is why ethical approval was not applicable. However, we obtained written consent to participate from all student participants and collected no data through which the participants could be identified. All authors have approved the manuscript and have given their consent to publish. Author contributions Adam Yousfi (adam.yousfi@helsinki.fi): Conceptualized the simulator and oversaw its development and produced 3D models. Conceptualized the study and recruited all study participants and supervised participant training and test task performance. Conducted all statistical analyses and wrote the manuscript. Ville Vasankari (ville.vasankari@hus.fi): Contributed key ideas to study design and supervised the study. Critically reviewed and contributed to the manuscript. Anni Pohjola (anni.pohjola@hus.fi): Contributed key ideas to study design and supervised the study. Critically reviewed and contributed to the manuscript. Emil Lintunen (emil.lintunen@gmail.com): Created simulator hardware and 3D models and reviewed and approved the manuscript. Elias Roos (elias.roos@gmail.com): Created simulator software and 3D models and reviewed and approved the manuscript. Martin Lehecka (martin.lehecka@hus.fi): Provided key conceptual contributions to both simulator design and development and study design. Critically reviewed and contributed to the manuscript. Availability of data and materials Our results are found in the manuscript file. However, we are willing to share the full data if contacted. We will share no data allowing identification of the participants. Disclosures This research was funded from the Helsinki University Hospital research funding. The research received no financial support from an organization beyond the authors' institutions. Funding This research was funded from the Helsinki University Hospital research funding. The research received no financial support from an organization beyond the authors' institutions. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.