A Novel Mixed-reality Simulator for Training Metacarpal Fracture Reduction and Kirschner Wire Fixation Authors: Akhil Nair, Rebecca Xu, Mark Shafarenko, Trinette Wright, Stephanie Williams, Jimmy Qiu, Daniel Lin, Stefan O.P. Hofer, Joseph Catapano, Blake Murphy Published: 2026 ABSTRACT Background:Metacarpal fracture reduction and Kirschner wire (K-wire) fixation are core competencies in hand surgery that are difficult to master. Surgical simulation using extended reality technologies has improved intraoperative performance among trainees in various surgical subspecialties.Methods:Imaging data from an uninjured patient were used as a reference to construct a 3D-printed hand. Wire channels and electromagnetic sensors were added to aid real-time tracking. Fracture bridges allowed introduction of fractures after single-piece printing, and the hand model was cast in silicone to reproduce surface anatomy. A scanned model of a K-wire driver was 3D-printed with movable parts, and an application was designed to track the positioning of the hand model, fracture sites, and the driver. Ten plastic surgeons and 8 residents evaluated the hand model simulator based on realism, representational accuracy, and training utility using a 5-point Likert scale for validation.Results:Printing accuracy showed a 3D printing error in Hausdorff distance below 0.13 mm. The target registration error for the fractures was below 1 mm. Qualitative evaluations of anatomical accuracy, clinical comparability, reduction realism of fractures, and the perceived value of simulator training each achieved a median score of 5. Percutaneous entry point selection, K-wire placement, fluoroscopy replicability, and training realism received a median score of 4. Overall, all plastic surgeons and residents unanimously recommended the simulator as an effective training tool.Conclusions:We introduce a validated mixed reality–integrated hand model simulator developed for training plastic surgery residents in metacarpal fracture reduction and fixation. Takeaways Question: What can be done to address a lack of operative training opportunities for plastic surgery trainees in metacarpal fracture reduction and fixation? Findings: Eighteen physicians rated the simulator model and found that it (1) recorded a low target registration error of less than 1 mm; (2) received high median scores for bone anatomy accuracy, clinical comparability of fractures, and training realism; and (3) was unanimously recommended by all evaluators as a training tool. Meaning: These findings validate the use of mixed reality and 3D-printing technology in the development of a training tool for plastic surgery residents. INTRODUCTION The traditional apprenticeship model of surgical training is deeply rooted in the Halstedian philosophy of “see one, do one, teach one.”1 Although the overall goal of acquiring incremental responsibility is expected, patient safety risks,2 underestimation of the complexities of various procedures,3 and the lack of guided practice limit its effectiveness. Subsequently, training models have experienced a pedagogical shift toward evidence-based methods, transforming the adage to “learn, see, practice, prove, do, and maintain.”4 The practice portion of the philosophy places strong emphasis on simulation-based mastery and maintenance of skills. This paradigm shift is a necessity in procedural training, as demonstrated by surveys assessing perceptions of preparedness among surgical residents. In a US national survey conducted in 2013, 38% of the respondents completing their general surgery training reported inadequate confidence in their operating skills, including 23% of graduating chief residents.5 A more longitudinal perspective from graduates of orthopedic residency programs in Ontario (1992–2020) reported that only 62.8% of participants felt ready to enter independent practice.6 Within a US plastic and reconstructive surgery resident cohort, although 87% felt that their operative experiences prepared them for practice, that number dropped to 76.6% and 52.4% when evaluating preparedness in hand and craniofacial surgery, respectively.7 Focusing specifically on core procedural competencies within hand surgery, 58%–72% of graduating residents from Canadian plastic surgery training programs (2004–2014) perceived themselves as insufficiently competent to perform 60% of commonly encountered procedures.8 To address these concerns regarding preparedness, there has been increased adoption of surgical simulation as a routine component of surgical training programs.9 Broadly defined, simulation includes a range of animal, cadaver, synthetic, and virtual models that seek to recreate the operative environment.9 Surgical simulation has been proven to improve resident performance in the operating room and competency scores across all training stages.10 In addition, structured simulation training has also been shown to accelerate learning curves among trainees,11,12 enhance skill acquisition,13 and reduce surgical adverse events.12 Amongst simulation models, virtual, augmented (AR), and mixed reality (MR) applications are of particular interest. Although virtual reality creates an independent digital environment for exploration through the use of wearable headsets or other computing devices, AR overlays digital information onto the real-world environment. MR systems, on the other hand, integrate real and virtual objects to coexist, allowing users to interact immersively with both the physical and virtual world. Because high-fidelity surgical simulation requires both visual and haptic feedback to maximize gains in skill acquisition, MR systems offer an ideal modality for training.14,15 Despite the increasing adoption of surgical simulation and AR or MR technology for trainee education, there is a paucity of literature pertaining to plastic surgery, with great prevalence in the spine, orthopedics, and general surgery fields.16–18 Given the technical complexity and variety of procedures performed by plastic surgeons, this specialty is uniquely poised to benefit from increased use of AR or MR simulations. Specifically, procedures with steep learning curves and requiring high visuospatial awareness would most likely benefit from high-fidelity simulations. One such procedure is Kirschner wire (K-wire) fixation of hand fractures, which requires subtle spatial coordination and tactile precision that can be challenging for novice learners.15 Multiple studies have attempted to create a surgical simulator for this purpose, but none have integrated an MR approach to provide an immersive training experience. Studies have either used anatomically accurate physical simulators to teach closed reduction and percutaneous pinning without any AR/MR components19–21 or incorporated surgical simulation adjuncts for a K-wire fixation task using GoogleGlass,22 a discontinued technology that is no longer commercially available. For the first time, we introduce an MR-integrated 3D-printed silicone hand and K-driver simulator with electromagnetic (EM) sensors to track real-time movement for surgical training of plastic surgery residents. METHODS Hand Model for Bennett and Fifth Metacarpal Fracture The following hand model was manufactured using the methodology described in Wright (Figs. 1A, B).23 A computed tomography scan of an uninjured, anatomically normal 52-year-old woman was segmented using ITK-Snap24 and then used as the reference for the model. The meshes of the hand were processed to create the 3D representation in Blender (Blender Foundation, Amsterdam, NL) and Meshmixer (Autodesk Inc., San Francisco, CA). Based on the inputs from plastic surgeons, the fifth metacarpal shaft and Bennett fracture at the base of the thumb were introduced. Skin meshes were subsequently adjusted to accommodate the new nonreduced thumb and fifth digit positions. A 3D-printable representation of the bones (Fig. 1A) was created using Autodesk Fusion 360 (Autodesk Inc, San Francisco, CA). Bridges (Fig. 1B) spanning the fracture sites were created to hold the pieces in the fractured position, allowing the bones to be printed as a single piece and later broken after silicone casting. Subsequently, EM sensors and wire channels (Fig. 2A) were added to the base of the thumb, the distal fracture piece of the fifth digit, and within the scaphoid and radius bones. To aid with tracking, 30 fiducials (Fig. 2A) were spread across the metacarpal of the thumb, distal fracture piece of the fifth digit, metacarpal of the index finger, trapezoid, and trapezium. The bones and the mold were 3D-printed with a Prusa i3 MK3S+ (Prusa Research, Prague) in polylactic acid. EM sensors from Northern Digital Inc. (Waterloo, Ontario) were secured in the wire channels for tracking spatial positioning, rotation, and orientation. Sequentially, FlexFoam-iT polyurethane foam was poured into the fractures, cured, and followed by soft-tissue casting in a 2-piece mold created from the skin mesh using silicone (Fig. 2). The fully formed hand model from the mold was trimmed at the seam lines, and fracture bridges were broken to create the Bennett and fifth metacarpal shaft fractures with independent freedom of motion. Figures 3A and B show the computer-aided design of hand bones in the mold and the fully assembled hand model, respectively. Hand model design specifics. A, Computer-aided design image of hand bones. B, Features of the hand model: purple indicates silicon adhesion holes; green indicates segmented bone; and blue indicates the bridge at the fracture site. Designing 3D-printed K-wire Driver The K-wire was modeled after the ConMed Hall 50 K-driver (CONMED Corporation, FL) (Fig. 3C). Each of the parts, including the knobs, lever handle, and hollowed main body of the driver, was designed and printed with polylactic acid as separate pieces to accommodate independent movement. To create a functional version of the driver, springs and elastics were added to create movable parts (Fig. 3D). The battery portion was not included in the design. We opted for a fixed K-wire tip made from a 1.1-mm diameter carbon fiber rod, which could withstand the pressure and torque applied by the user while not introducing additional artifacts to the EM field. Simulator Software A simulator software program was developed in-house for real-time tracking of the hand model, K-wire driver, and their mesh representations. Data from EM sensors were used to update the position and orientation of the hand model, fractured bones, and K-wire driver in the renderer. The software tracks the advancement of the virtual K-wire and can distinguish between intentional versus accidental activation. Additionally, contacts between the virtual K-wire and the metacarpal bone are recorded through the computation of K-wire line intersection with underlying model meshes. Audiovisual feedback indicates driver engagement when the lever and button are pressed together. A color change to green is used to convey that the K-wire has contacted the bone. Finally, the simulator also has the capability to capture virtual radiographs in anterior-posterior, lateral, and oblique orientations. (See Video 1 [online], which demonstrates the simulator and shows a participant performing reduction of the Benett fracture and capturing virtual radiographs.) (See Video 2 [online], which demonstrates the simulator and shows a participant performing K-wire fixation using the 3D-printed driver and capturing virtual radiographs.) Video 1.Demo of the simulator which shows a participant performing reduction of the Benett’s fracture and capturing virtual radiographs.GOX.0000000000007836video1.mp4Kaltura Demo of the simulator which shows a participant performing reduction of the Benett’s fracture and capturing virtual radiographs. Video 2.Demo of the simulator which shows a participant performing K-wire fixation using the 3D printed driver and capturing virtual radiographs.GOX.0000000000007836video2.mp4Kaltura Demo of the simulator which shows a participant performing K-wire fixation using the 3D printed driver and capturing virtual radiographs. Validation of the Model Ten practicing plastic surgeons and 8 plastic surgery residents at various stages of training evaluated the model and the K-wire fixation simulator. All participants were instructed to perform closed reduction and K-wire fixation on the Bennett fracture, followed by completion of an anonymous 10-question survey evaluating hand model realism, clinical realism of reduction and fixation, and educational utility of the simulator. Additional space was provided within the survey for respondents to suggest improvements to the MR simulator. Responses were recorded on a 5-point Likert scale (5, strongly agree; 4, agree; 3, neutral; 2, disagree; and 1, strongly disagree). Results of the responses were further analyzed for median scores, and a Likert score distribution plot was generated using RStudio (version 4.4.2, R Foundation for Statistical Computing, Vienna, Austria). RESULTS We were successfully able to create an anatomically accurate hand model with EM sensors, coupled with the 3D-printed K-wire model, to be used for training in both fracture reduction and K-wire fixation. The 3D-printing reproducibility was confirmed by printing 4 partial thumbs and index phalanges, followed by scanning them on Siemens Cios Spin cone beam computed tomography and comparing their segmentations to the original meshes (Table 1). Errors in 3D-printing were calculated using the Hausdorff distance function in MeshLab25 and were measured to be 0.11 and 0.12 mm for the index finger and thumb, respectively, even with the use of 2 different 3D printers.23 The error rates were comparable between 2 separate prints from the same 3D printer versus printing on a different 3D printer. The mean fiducial registration errors (Table 2) across 3 hands were 0.55, 0.69, and 0.74 mm for the hand, thumb, and fifth digit, respectively.23 The target registration error was measured to be less than 1 mm around the fracture locations on the thumb and the fifth metacarpal bone (Fig. 4). Table 1.3D-printing Errors, Calculated as the Mean Hausdorff Distance (mm), for 2 Partial Prints Each of the Index Finger and Thumb Produced on 2 Different 3D Printers and Compared With the Mesh Generated From the Original Computed Tomography Hand Scan (F1, F2: 2 Prints From 3D Printer A; P1, P2: 2 Prints From 3D Printer B)Partial PrintMean Hausdorff Distance ± SD, mmRange (Min.–Max.), mmF1 index0.107 ± 0.084<0.01–0.567F2 index0.094 ± 0.079<0.01–0.590P1 index0.118 ± 0.097<0.01–0.672P2 index0.134 ± 0.103<0.01–0.658F1 thumb0.1210 ± 0.095<0.01–0.631F2 thumb0.105 ± 0.080<0.01–0.611P1 thumb0.116 ± 0.086<0.01–0.562P2 thumb0.139 ± 0.103<0.01–0.570Values are reported in millimeters; ranges with minima less than 0.01 mm are reported as less than 0.01. 3D-printing Errors, Calculated as the Mean Hausdorff Distance (mm), for 2 Partial Prints Each of the Index Finger and Thumb Produced on 2 Different 3D Printers and Compared With the Mesh Generated From the Original Computed Tomography Hand Scan (F1, F2: 2 Prints From 3D Printer A; P1, P2: 2 Prints From 3D Printer B) Values are reported in millimeters; ranges with minima less than 0.01 mm are reported as less than 0.01. Table 2.FRE (Measured in mm) for Different 3D-printed ModelsSpecimenFRE, mmMean ± SD, mmRange (Min.–Max.), mmHand model 10.630.55 ± 0.270.24–0.78Hand model 20.24Hand model 30.78Thumb 10.650.686 ± 0.040.65–0.73Thumb 20.68Thumb 30.73Fifth digit 10.70.736 ± 0.0350.7–0.77Fifth digit 20.74Fifth digit 30.77FRE, fiducial registration error. FRE (Measured in mm) for Different 3D-printed Models FRE, fiducial registration error. Among plastic surgeons, survey responses regarding surface anatomy realism received a median score of 4.5, whereas bone anatomy and clinical accuracy of Bennett fracture both demonstrated a median score of 5 (Table 3). The Likert score distribution (Fig. 5A) shows that all responses regarding bone anatomy and clinical accuracy of Bennett fracture ranged from agree to strongly agree, with no respondents selecting disagreement options. In contrast, responses regarding surface anatomy realism ranged from neutral to strongly agree. Responses regarding reduction manipulation mimicking clinical situations yielded a median score of 5, whereas K-wire entry point selection and placement and MR images replicating fluoroscopy in the operating room achieved a median score of 4. In addition, responses evaluating the overall applicability of the training model to provide realistic training and its value for Bennett fracture K-wire fixation training achieved median scores of 4 and 5, respectively. A major proportion of respondents consistently chose agree or strongly agree in all those instances except K-wire placement realism, where 40% of responses were neutral. Among plastic surgery residents (Fig. 5B), the responses to the survey questions followed a similar pattern with minor variations noted in a couple of instances. The residents recorded a median score of 3.5 in response to the question evaluating whether the reduction manipulations of the model approximated the clinical situation, where 50% of the responses were neutral. A median score of 4.5 was also recorded for questions pertaining to bone anatomy accuracy and the value of the simulator for Bennet fracture training, noting a 0.5 reduction in median scores from plastic surgeons (Fig. 5). Finally, the survey question assessing whether plastic surgeons and residents would personally recommend this simulator as a training tool was rated with a median score of 5, with both groups choosing either agree or strongly agree in all responses. Table 3.Median Likert Scale Scores Based on Model Evaluation by Plastic Surgeons and ResidentsQuestionMedian Score Plastic Surgeons (n = 10)Median Score Plastic Surgery Residents (n = 8)Surface anatomy realism4.54.5Bone anatomy accuracy54.5Clinical accuracy Bennett fracture55Reduction manipulation replicates clinical situation53.5K-wire entry point selection simulation realism44K-wire placement in fracture simulation realism44MR images replicability to fluoroscopy44Applicability of the simulator for providing realistic training44Overall value of the simulator for Bennett fracture training54.5Personal recommendation of the simulator as a training tool55 Median Likert Scale Scores Based on Model Evaluation by Plastic Surgeons and Residents DISCUSSION With the advancement of digital technologies, innovative simulation tools are constantly being incorporated into surgical training due to their interactive and immersive nature. Despite growing use in other surgical subspecialties,26 the underuse of virtual tools in plastic surgery training10 could stem from difficulty in convincingly replicating soft-tissue realism, cost, and the nuanced tactile feedback necessary for surgical fidelity. To this end, we have developed a hand model that looks and feels similar to a real hand, replicates fractures as regularly seen in clinical settings, and also allows for virtual K-wire fixation that is accurate within a distance of 1 mm. Survey responses from 10 practicing plastic surgeons and 8 plastic surgery residents demonstrated that the hand model was not only perceived as highly accurate in replicating anatomical and fracture characteristics, but was also unanimously recommended as a training tool. In addition, for a 1-time upfront cost of a $1930 computer-aided design, the hand model can be used in hundreds of training sessions without incurring any meaningful damage, because K-wire insertion is simulated with MR. Furthermore, improvements in guidewire placement accuracy, intraoperative efficiency, and reduced need for fluoroscopic images have already been demonstrated with the use of hybrid reality EM-tracked simulators in orthopedic surgery.27 Similarly, we would expect trainees to improve their intraoperative accuracy in K-wire placement with the use of our simulator, given its tracking accuracy and realism. In accordance with the shift toward competency-based curricula, the integration of objective scoring systems is essential for tracking trainee progress. Hybrid reality systems have demonstrated their effectiveness, with studies using validated tools such as Objective Structured Assessment of Technical Skills showing significant improvement in scores and reduced operative times in basic laparoscopic procedures.28 Such integration of objective scoring and simulation training would allow for tracking improvement in skill acquisition without compromising on patient safety. Future directions will include piloting this simulator among plastic surgery residents to track both skill acquisition and the transferability of these skills into the operating room. Predominant suggestions from the evaluators to improve the model included haptic feedback for K-wire piercing and advancement, an option to realign the driver after “piercing” the skin, and attachment of the hand to an arm to further replicate operating room restrictions. We chose not to include the haptic feedback option at this stage, considering its engineering challenges, focusing instead on our primary goal of creating a system that enhances user spatial awareness and provides a realistic simulation to master the procedure. We plan to include all these suggestions in the future versions of our MR simulator. In conclusion, we introduce and validate a novel MR hand model simulator for fracture reduction and K-wire fixation, offering a practical solution for acquiring a skill that is deceptively challenging to master. Widespread adoption of validated simulators will enhance the plastic surgery training program by providing consistent and accessible skill acquisition while contributing to improved patient safety. Hand model design specifics. A, Green indicates the wire channel in the thumb; blue indicates fiducial points for EM registration. B and C, Two-part mold for 3D bone and soft-tissue casting Hand model design specifics. A, Computer-aided design image of the hand bones within the mold. B, Assembled hand model. C, Reference model of the K-wire driver. D, Assembled K-wire driver with carbon fiber tip. Visualization of target registration errors for the thumb (A), hand (B), and finger (C). All distances are recorded in millimeters. Likert score distribution of the 10-question survey. A, Plastic surgeons (n = 10). B, Plastic surgery residents (n = 8). DISCLOSURES The authors have no financial interest to declare in relation to the content of this article. This project was funded by the Munk Innovation Fund. ETHICAL APPROVAL Institutional research ethics board approval was obtained from the University Health Network (nos: 20-5241 and 24-5294).