Augmented Reality-Assisted System for Tooth Carving Training in Dental Education

Summary

The Augmented Reality-Assisted Tooth Carving System (AR-ATCS) recognizes 3D object markers and superimposes a virtual tooth model onto a real wax block, providing a three-dimensional visual reference to assist novice dental students during tooth carving.

Abstract

The goal of this protocol is to enhance dental anatomy education by developing and applying an augmented reality-assisted tooth-carving system (AR-ATCS). Tooth morphology is a fundamental component of dental education, requiring students to master complex three-dimensional (3D) anatomical structures and functional landmarks that are essential for clinical procedures such as restorative dentistry and endodontic treatment. Traditionally, these skills are acquired through wax-block carving exercises; however, this approach is often limited by low instructional efficiency and a substantial cognitive gap when students attempt to translate two-dimensional (2D) textbook illustrations into three-dimensional physical structures. To address these challenges, this protocol describes integrating augmented reality (AR) technology into manual tooth-carving training. The system uses three-dimensional object tracking and computer vision algorithms to superimpose a high-fidelity virtual 3D tooth model onto the physical workspace. By recognizing dedicated tracking markers with a mobile device camera, the system projects a digital guide at the correct scale and orientation directly onto the wax block. This virtual-to-physical alignment enables students to visualize the target tooth morphology in real time and identify deviations in their carving progress as they occur. The protocol provides a detailed workflow for establishing the AR environment, aligning virtual models with physical materials, and using the system as a self-guided training tool. By delivering interactive, multi-angle visual feedback, the system reduces dependence on continuous instructor demonstrations and minimizes the impact of irreversible carving errors. Overall, this AR-assisted approach provides a supplementary visual guidance tool for traditional dental anatomy training by supporting novice students during tooth carving and offering a more efficient framework for practicing tooth morphology.

Introduction

Dental morphology is a core component of dental education, aiming to help students systematically understand tooth anatomical morphology and its three-dimensional structural characteristics^1,2. Tooth carving practice, as an important part of dental morphology teaching, not only reflects students’ mastery of tooth anatomy but also trains their manual dexterity and operative skills^3,4. It provides an essential foundation for clinical practice, particularly in prosthodontic treatment^2,5.

Traditional tooth carving training is usually conducted under the guidance of instructors, in which students carve the anatomical morphology of anterior and posterior teeth using wax or soap blocks^3,5. However, the imbalance between the number of instructors and students makes it difficult for teachers to provide sufficient individualized guidance within a limited teaching period^4,5. In addition, students often experience considerable cognitive difficulties when transforming two-dimensional images into three-dimensional forms, which may reduce both learning efficiency and teaching effectiveness^2,5,6.

In recent years, with the rapid development of digital technologies, three-dimensional printing, augmented reality, virtual reality, and other digital approaches have gradually been applied in medical anatomy education and dental laboratory teaching^6,7,8,9. These technologies overcome the limitations of traditional two-dimensional atlases, physical model observation, and teacher demonstration in spatial presentation, providing students with a more intuitive, dynamic, and three-dimensional learning experience^5,6,9.

Augmented reality is a visualization technology that superimposes computer-generated virtual information onto the real environment in real time, thereby achieving virtual-real fusion and human–computer interaction. By supplementing three-dimensional data and spatial information in real-world scenarios, AR can expand the observer’s perception of complex spatial structures^10,11. This feature makes AR particularly suitable for practical teaching scenarios that require spatial understanding and hand–eye coordination. At present, AR technology has been preliminarily applied in anatomy education and has received positive feedback^11,12,13. Beyond educational settings, AR has also been translated into clinical dentistry. For example, Macrì et al. reported the use of AR-assisted surgical exposure of an impacted tooth, demonstrating that AR-based spatial registration and real-time virtual-real overlay can be applied in oral surgical navigation¹⁴. This clinical application provides additional support for the introduction of AR-guided spatial visualization into dental morphology training.

In this protocol, AR technology was used to assist tooth-carving training by providing students with a continuous, intuitive three-dimensional visual reference during practice, which may serve as a supplement to conventional instructor guidance. For this purpose, an AR-assisted tooth-carving system was jointly developed and primarily consists of AR software, a high-definition camera, a wax-block carrier base, and four separately fabricated three-dimensional object-tracking markers. The markers were assembled onto the carrier base, which holds the wax block in a predefined position. During operation, the system captures images of the physical working area through the camera. After a three-dimensional object marker on the base is recognized, the virtual tooth model is superimposed in real time onto the corresponding position on the wax block in the display interface, thereby providing students with a visual reference for carving.

Based on this system, a single-sample validation was performed. A student who had completed the dental anatomy course completed the tooth-carving procedure under the guidance of the proposed AR-assisted protocol. The carved tooth was then scanned and processed as a three-dimensional model, and an objective morphological deviation analysis was conducted by comparing it with the standard tooth model. The preliminary single-sample results suggested the feasibility of using the AR-ATCS as an auxiliary tool for dental morphology training.

Protocol

1. Establishment of the standard digital tooth model

NOTE: In this protocol, the left mandibular first molar was used as the target tooth to demonstrate the feasibility of the AR-assisted tooth-carving workflow. For other tooth positions, the workflow can be applied after importing the corresponding standard tooth digital model and redefining its spatial relationship with the three-dimensional object marker. This step takes approximately 10 min.

1. Chamber scanning of the standard tooth model
   1. Prepare the standard tooth model.
      NOTE: In this protocol, use a standard tooth model compatible with a 25 mm × 25 mm × 50 mm wax block.
   2. Apply a thin and uniform layer of scanning powder or anti-reflective spray to the surface of the standard tooth model to reduce reflection-related scanning errors.
   3. Place and secure the prepared standard tooth model onto the holder of the chamber-type three-dimensional scanner (equipped with dual 24 MP cameras and scanning accuracy <10 µm). Ensure that the model is stable and that the target scanning area is fully exposed (Figure 1A).
   4. Open the 3D scanning software, create a new order, and name it according to the target tooth position. Confirm the scanning settings and click on Start Scan.
   5. After scanning, inspect the integrity of the digital model. If no obvious defects are observed, export the scanned model in STL format and set the unit to millimeters (Figure 1B).
2. Segmentation and processing of the scanned tooth model (Figure 1C)
   1. Open 3D scan data processing software.
   2. Click on File > Open, select the STL file of the standard tooth model, and click on Open.
   3. In the unit selection window, select Millimeters and click on OK.
   4. Click on Polygons > Trim > Trim with Plane. Manually adjust the Position of the cutting plane to the planned root sectioning level.
   5. Confirm the cutting region, click on Intersect Plane > Delete Selection > Close Intersection > OK to remove the excess root portion and close the open boundary, and obtain a watertight standard digital tooth model.
   6. Click on File > Save and choose STL (binary) File (*.STL) in the File Type list.
3. Construction of the standard tooth model base (Figure 1D)
   ​NOTE: This step is used to construct a regular base for the standard digital tooth model, thereby establishing a stable spatial relationship among the standard model, wax block, and AR tracking base. The exact names of menu options may vary slightly among software versions.
   1. Open the 3D modeling software.
   2. Click on File > Open and select the STL file of the standard digital tooth model.
   3. Click on Options, import as Solid Body or Surface Body, and select Create mesh bodies bounded by single faces > Group facets into faces, and click on OK and Open, so that the triangular facets at the bottom of the tooth model can be recognized as a relatively complete reference face.
   4. Create the center point of the model base. Select the bottom surface of the tooth model, click on Features > Point > Center of Face > OK to generate the center point (named Point1).
   5. Choose the bottom surface of the tooth model as the sketch plane and click on Sketch > Center Rectangle, draw a rectangle centered at Point 1, and click on Smart Dimension to set the rectangle to 25 mm × 25 mm. Click on Exit Sketch (named Sketch 1).
   6. Select Sketch 1. In the Features tab, click on Extruded Boss/Base. Under Direction 1, select Blind, set the Depth to 20 mm, and click on OK to generate the standard tooth model base.

2. Design of the wax-block holder and AR tracking base

NOTE: The three-dimensional object-tracking markers and the AR tracking wax-block base are customized components. Their design may vary depending on the design purpose and modeling strategy. Therefore, the base described in this protocol is not the only possible design and may be modified, provided that wax-block fixation, marker recognition, and handling comfort are maintained. This step takes approximately 30 min.

1. Design of the wax-block recess (Figure 2A,B)
   1. Using the 3D modeling software, click on New > Part > OK on the home page.
   2. Choose the Top Plane and click on Sketch. Click on Center Rectangle and draw the initial sketch centered at the Origin.
   3. Click on Smart Dimension to set the rectangle to 40 mm × 40 mm, click on Exit Dimension (named Sketch 1).
   4. Choose Sketch 1, click on Features > Extruded Boss/Base, under Direction 1, choose Blind, and set the Depth to 32.5 mm to create a 40 mm × 40 mm × 32.5 mm cuboid (named Boss – Extrude 1).
   5. Choose the top surface of the cuboid, click on Sketch > Center Rectangle, and create the initial sketch centered at the Origin. Then click on Smart Dimension and set the rectangle size to 25.2 mm × 25.2 mm (named Sketch 2). This size accommodates a 25 mm × 25 mm wax block with an assembly tolerance of approximately 0.2 mm.
   6. Choose Sketch 2, click on Features > Extruded Cut, and set the cut depth to 20 mm to create the wax-block recess.
2. Design of negative sockets for the three-dimensional object tracking markers
   1. Choose one vertical surface of the cuboid and click on Sketch > Center Rectangle, draw a rectangle on the chosen face, click on Smart Dimension, and set it to 5.2 mm × 5.2 mm (named Sketch 3).
   2. Click on Features > Extruded Cut, choose Sketch 3, and set the cut depth to 5 mm.
   3. Repeat the above steps to create sockets on the four vertical surfaces. For the socket on the bottom surface, set the sketch size to 10.2 mm × 10.2 mm and the cut depth to 10 mm.
   4. Click on Save and save the wax-block holder as an SLDPRT file.
3. Design for base stability and carving comfort
   1. Using the 3D modeling software, click on New > Part > OK on the home page.
   2. Choose the Top Plane and click on Sketch > Circle, draw an 80 mm-diameter circle centered at the Origin (named Sketch 1).
   3. Choose Sketch 1, click on Features > Extruded Boss/Base, and set the depth to 5 mm to create a circular stabilizing plate.
   4. Click on Right Plane and draw an isosceles right triangle with the Origin as the midpoint of the hypotenuse, a side length of 40 mm, and a 45° angle to create a sketch (named Sketch 2).
   5. Choose Sketch 2, click on Features > Extruded Boss/Base, choose Mid Plane under Direction, and set the depth to 40 mm to generate a 45° inclined structure (named Boss-Extrude 2).
   6. Choose the inclined surface, click on Sketch> Center Rectangle, draw a 10 mm × 10 mm center rectangle, click on Extruded Boss/Base, and set the depth to 10 mm to create the supporting connector to the base.
   7. Click on Save and save the stabilizing support structure as an SLDPRT file.
4. Design of the three-dimensional object tracking markers
   ​NOTE: The marker should have a rigid structure, stable surface features, and sufficient geometric complexity to improve the stability of three-dimensional object recognition and tracking.
   1. Using the 3D modeling software, click on New > Part > OK on the home page.
   2. Choose the Top Plane and click on Sketch > Polygon, draw a hexagon centered at the Origin, click on Smart Dimension, set the side length to 7.5 mm (named Sketch 1).
   3. Choose Sketch 1, click on Features > Extruded Boss/Base, and set the depth to 10 mm.
   4. Choose the bottom face, click on Sketch> Center Rectangle, draw a 5 mm × 5 mm center rectangle (named Sketch 2).
   5. Choose Sketch 2, click on Features > Extruded Boss/Base, and set the depth to 5 mm to form a pin that fits into the negative socket of the base.
   6. Use the Draft function with the top face as the Neutral Plane. Select each side surface as the Draft Plane, and apply different draft angles to each surface to increase the asymmetry of the geometric features. Adjust the draft angles based on recognition performance, while avoiding complete symmetry across all side surfaces.
   7. Choose the top surface of the marker, click on Sketch > Text, and enter a text label to add local visual features and assist recognition (named Sketch 3).
   8. Choose Sketch 3, click on Features > Extruded Boss/Base, and set the extrusion depth to 2 mm.
   9. Repeat the above steps to create the four side markers corresponding to the mesial, distal, buccal, and lingual surfaces.
   10. Click on Save and save each marker as an SLDPRT file.
5. Assembly of the base and AR markers and confirmation of spatial relationships
   1. Using the 3D modeling software, click on New > Assembly > OK on the home page.
   2. Click on Insert Components, import the AR tracking wax-block base, stabilizing support structure, and all three-dimensional object tracking markers.
   3. Click on Mate function and apply Coincident, Parallel, Perpendicular, or other mating relationships to define the spatial positions between the markers and the base.
   4. Click on Save As, select STL format, and click on Options in the window.
   5. Confirm Save all components of an assembly in a single file is not selected, click on OK > Save, export the individual component STL files separately for model import and spatial relationship setup in the AR software.

3. 3D printing and post-processing of the models

NOTE: Using the printer and printing parameters described in this protocol, printing a complete set of models takes approximately 70 min.

1. Material selection and printer setup (Figure 2C,D)
   1. Use matte gray polylactic acid (PLA) filament for printing to reduce reflection and facilitate recognition by the AR camera.
   2. Use an enclosed FDM 3D printer to control the printing environment and improve printing stability. Adjust the parameters to match the printer and filament characteristics.
   3. Import the STL model into the slicing software. Determine the optimal orientation by moving, rotating, and selecting the printing surface to reduce printing errors.
   4. Configure the slicing parameters using the slicing software. The parameters may be adjusted according to the specific printer and material.
2. Recommended slicing parameters
   1. Set the nozzle diameter to 0.4 mm, the nozzle temperature to 220 °C, and the build plate temperature to 55 °C. Use a textured PEI build plate.
   2. Set the layer height to 0.2 mm, the infill density to 15%, and the printing speed to 50 mm/s for the first layer, 200 mm/s for the outer wall, and 300 mm/s for the inner wall.
   3. Enable automatic tree supports when overhanging structures are present. Omit support structures if no overhangs are present. Enable cooling.
   4. Calibrate the printer before printing and inspect the filament color and surface condition.
3. Post-processing of printed models
   1. After printing and cooling, remove the build plate and gently bend it to detach the model naturally, thereby protecting the surface texture of the build plate.
   2. Carefully remove the support structures using needle-nose pliers. After removing the supports, inspect the model for any integrity issues or deformation. If the support-contact areas are rough, gently polish them with sandpaper or a file to improve base stability.

4. AR software setup and virtual-real registration

NOTE: In this protocol, augmented reality (AR) software was used to import the standard digital tooth model and generate the virtual-real overlay. A high-definition RGB camera was used to capture 1080p RGB images at 30 fps. The camera's depth-sensing function was not used for tracking or registration (Figure 3A,B). Importing the virtual models into the software and setting the model color and transparency took less than 1 min. The registration process took less than 5 s, and real-time navigation was achieved immediately after successful registration.

1. Hardware preparation
   1. Turn on the computer and connect the high-definition camera via a USB 3.0 port. Confirm that the camera is recognized by the system.
      ​NOTE: The camera calibration parameters were predefined during software development and used for AR registration. No additional calibration was required during subsequent use. For other AR systems, repeated camera calibration should be performed according to the system requirements.
2. Model import and display settings
   1. Open the AR software.
   2. Click on Import STL, select the folder containing the STL files of the standard digital tooth model and AR marker models, and import the required models.
   3. Set the virtual tooth model to white to provide high contrast with the red wax block, and set the transparency to 0.4 to facilitate observation of the relationship between the carving instrument and the boundary of the tooth model.
3. Starting AR navigation and registration
   NOTE: During registration, avoid occluding the marker with the hand or carving instruments. If the system cannot recognize the marker, remove the obstruction and readjust the base so that the marker re-enters the camera view clearly.
   1. Select the AR marker and click on Virtual Navigation > Camera Navigation. In the pop-up window, select the camera to be used and start AR navigation.
   2. Place the physical three-dimensional object tracking marker at the center of the camera view, maintaining a working distance of approximately 25 cm. Use a non-reflective background and avoid strong direct illumination in the operating area to maintain stable marker recognition.
   3. Adjust the camera angle so that the optical axis is approximately perpendicular to the marker surface or forms an oblique viewing angle of approximately 30°–45°, allowing the virtual marker outline in the software to overlap the physical marker as closely as possible in size, orientation, and position.
      ​NOTE: When the virtual marker closely overlaps the physical marker, the system automatically recognizes it and completes registration, enabling real-time tracking. The virtual marker outline changes from red to green, and the virtual tooth model is then displayed according to its predefined spatial relationship with the marker and is stably superimposed at the corresponding position on the wax block.
   4. Click on Parameter Settings to adjust the tracking quality threshold.
      NOTE: This threshold is used to determine the matching quality between the virtual model and the real scene; registration is considered successful only when feature matching and localization confidence reach the preset threshold. In this system, the tracking quality threshold was set to 0.7 based on repeated preliminary tests. This value was selected because it allowed rapid recognition while maintaining stable real-time tracking and virtual-real registration. Lower thresholds allowed easier marker registration but reduced registration accuracy, whereas higher thresholds increased the difficulty and time required for marker registration.

5. AR-assisted tooth carving procedure

1. Insert the 25 mm × 25 mm × 50 mm wax block into the recess of the AR tracking wax-block base. Ensure that the bottom of the wax block is fully seated against the recess floor and that the wax block is stable.
2. Start AR navigation. Determine the initial cutting area and direction based on the overlay between the virtual tooth model and the wax block. Rotate the base slowly to observe the virtual tooth model from different viewing angles.
3. Observe the virtual tooth model from the viewing direction corresponding to the recognized three-dimensional object tracking marker and its adjacent surfaces. Avoid relying on the opposite surface, as it may not be displayed stably when only a single marker is recognized.
4. Rotate the wax-block holder to expose the mesial, distal, buccal, or lingual three-dimensional object tracking marker as needed. In Model Settings, select the corresponding marker to switch the viewing surface.
5. Adjust the color and transparency of the virtual tooth model in Model Settings according to the stage of observation.
6. Remove any hand or instrument that occludes the marker. If the system fails to recognize the marker due to occlusion or rapid base rotation, slowly readjust the base position until the marker is recognized again, then continue carving (Figure 3C,D).

Results

The AR-ATCS successfully recognized the three-dimensional object-tracking marker on the base and stably superimposed the virtual tooth model onto the corresponding position on the wax block in real time. The virtual model accurately represented major anatomical structures, morphological boundaries, and surface contours, including the cusps, fossae, and grooves, marginal ridges, axial surface contours, and crown outline. When the base was rotated slowly, and the tracking marker remained unobstructed, stable registration between the virtual tooth model and the wax block was maintained. This enabled visualization of the target tooth morphology from multiple perspectives, including the buccal, lingual, mesial, distal, and occlusal views. These findings demonstrate that the system provides a continuous three-dimensional visual reference that facilitates identification of local morphological deviations during tooth carving.

When the tracking marker was partially occluded by the operator's hand or carving instruments, or when the base was rotated too rapidly, marker recognition was interrupted, resulting in loss of the virtual model overlay. Repositioning the base and removing the obstruction restored marker recognition and re-established stable virtual-real registration, allowing carving to resume.

A two-dimensional image-plane overlay deviation analysis was performed to evaluate the virtual-real overlay error. For each registration trial, the camera was fixed, and the tooth-carving base with the object-tracking marker was moved to the center of the camera's field of view. When the virtual marker outline approximately overlapped the physical marker outline, the system automatically completed registration. A two-dimensional overlay image was then captured. Four corner landmarks of the tooth-carving base were selected from each captured image, and the corresponding physical and virtual landmark coordinates were recorded (Figure 4). The Euclidean distance between each physical–virtual point pair was calculated as the two-dimensional overlay deviation. For each surface, 10 repeated registration trials were performed, and four corresponding physical–virtual landmark pairs were measured in each trial. Thus, 40-point pairs were analyzed per surface, yielding a total of 200-point pairs across five surfaces. For the buccal, distal, mesial, and lingual surfaces, registration was performed using the object-tracking marker corresponding to each surface. For the occlusal surface, registration was first completed using markers on the other surfaces, and the base was then rotated to obtain the occlusal-view overlay image. This additional rotation may have contributed to the larger deviation observed in this view. The overall mean two-dimensional overlay deviation was 0.679 ± 0.381 mm, with an RMS deviation of 0.778 mm. The occlusal surface showed the largest deviation and variability, whereas the buccal, distal, mesial, and lingual surfaces showed smaller overlay deviations (Table 1).

The carving result presented here was obtained from a single student who had recently completed a dental anatomy course and completed the tooth carving procedure within 2 hours under continuous guidance from the AR-ATCS; therefore, the quantitative deviation analysis represents single-sample data. Three-dimensional deviation analysis was performed by comparing the scanned carved tooth model with the standard tooth model. Using a predefined tolerance of ±0.5 mm, 52.26% of the surface points fell within this range. The signed mean deviation was 0.1025 mm, and the root mean square (RMS) deviation was 0.7434 mm. The maximum positive and negative deviations were +1.9275 mm and −1.9272 mm, respectively. The color-coded deviation map indicated that the greatest morphological discrepancies were localized primarily to the occlusal surface and selected axial contour regions (Figure 5). These deviation values represent the cumulative discrepancy between the final carved tooth model and the standard tooth model after three-dimensional scanning and model alignment, rather than isolated measurements of carving error, AR registration error, or scanning error.

These preliminary single-sample results support the technical feasibility of the AR-ATCS workflow. During the observed tooth-carving procedure, the system maintained a real-time overlay of the virtual tooth model onto the physical wax block, providing a three-dimensional reference for tooth morphology and related anatomical features.

Figure 1: Preparation of the standard wax tooth digital model. (A) Three-dimensional scanning of the standard wax tooth. (B) Initial scan data of the standard wax tooth and surrounding structures. (C) Processed the standard tooth digital model after trimming and surface optimization. (D) The standard tooth model was aligned with the virtual wax block base to define its spatial position in the AR system. Please click here to view a larger version of this figure.

Figure 2: Design and fabrication of the wax-block carrier base. (A) Three-dimensional design model of the wax-block carrier base and the separately designed three-dimensional object tracking marker. (B) Spatial relationship setup between the standard tooth digital model and the wax-block carrier base. (C) The 3D-printed wax-block carrier base with the separately printed object tracking marker. (D) The wax block is mounted on the 3D-printed carrier base for subsequent AR-assisted tooth carving.

Figure 3: Workflow of augmented reality-assisted tooth carving. (A) Interface of the augmented reality navigation software showing the imported standard tooth digital model and a three-dimensional object-tracking marker. (B) After recognizing the three-dimensional object marker on the base, the system superimposes the virtual standard tooth model onto the corresponding position of the physical wax block in real time. (C) Initial tooth carving is performed under AR guidance using the virtual tooth model as a three-dimensional morphological reference. (D) Fine adjustment of local anatomical structures, including cusps, fossae, and grooves, marginal ridges, and axial surface contours, is performed under AR guidance. Please click here to view a larger version of this figure.

Figure 4: Measurement of two-dimensional image-plane overlay deviation after AR registration. After automatic AR registration, a two-dimensional overlay image was captured. P1–P4 indicate the four corner landmarks selected on the physical tooth-carving base, and P1′–P4′ indicate the corresponding landmarks selected on the virtual overlay. Please click here to view a larger version of this figure.

Figure 5: Three-dimensional deviation analysis of the carved tooth model compared with the standard tooth model. Three-dimensional deviation maps are shown from the buccal view (A), lingual view (B), occlusal view (C), mesial view (D), and distal view (E). Please click here to view a larger version of this figure.

Surface	Point pairs, n	Mean ± SD (mm)	RMS (mm)	Min (mm)	Max (mm)
Buccal surface	40	0.711 ± 0.226	0.745	0.235	1.174
Distal surface	40	0.489 ± 0.156	0.513	0.146	0.913
Mesial surface	40	0.510 ± 0.207	0.549	0.194	0.856
Lingual surface	40	0.540 ± 0.262	0.599	0.181	1.098
Occlusal surface	40	1.145 ± 0.494	1.245	0.27	2.401
Overall	200	0.679 ± 0.381	0.778	0.146	2.401

Table 1: Two-dimensional image-plane overlay deviation of AR virtual-real registration.

Discussion

Dental anatomy education has evolved over time, adopting a classical teaching model that combines theoretical instruction with practical training. An accurate understanding of tooth anatomical morphology is an important foundation for oral clinical practice, particularly in procedures related to prosthodontic treatment, endodontic treatment, occlusal reconstruction, and implant restoration. Therefore, researchers have continuously explored new teaching methods and technologies to support the establishment of anatomical knowledge systems. Previous studies have mainly improved dental anatomy education by optimizing teaching methods and exploring the applicability of different carving materials. With the development of digital technologies, three-dimensional scanning, three-dimensional reconstruction, 3D printing, virtual reality, and augmented reality have gradually expanded from clinical applications to medical education, providing new tools for students to understand complex three-dimensional structures¹⁵. In the future, as artificial intelligence and mobile computing develop, these technologies are expected to be integrated into routine teaching in a lower-cost, more convenient manner.

In this system, the standard tooth model can be reused. The three-dimensional object tracking markers can be designed in 3D modeling software, and their spatial relationships can be determined before being directly imported into the AR software. The entire workflow can then be completed by 3D-printing the tooth-carving base. It is worth noting that using 3D-printed tooth models as physical references for tooth-carving training has also shown educational value, with advantages such as low cost, intuitive morphology, and repeated observation. However, AR technology offers real-time visual navigation. It can provide a real-time, rotatable, and transparency-adjustable three-dimensional visual reference in the actual wax block carving environment¹³. As the technology continues to develop, AR is expected to be applied to education in a lower-cost, simpler, and more convenient form. Therefore, AR-assisted teaching has considerable potential for further development.

For AR systems, visualization quality, tracking stability, and registration accuracy are key technical factors that determine the reliability of virtual-real overlay and the user experience¹⁶. Previous studies have explored the use of AR systems for tooth carving education, but most have adopted image tracking^5,13. In contrast, the present system uses three-dimensional object tracking. Image marker tracking offers simple preparation, low cost, and easy deployment, and it can provide stable recognition when the marker is clearly visible and within an appropriate viewing angle. During tooth carving, the wax block often needs to be rotated so that the operator can observe the target tooth morphology from multiple angles. Under these conditions, an image marker may not remain fully visible within the camera field of view, which can interrupt marker recognition and limit the continuity of real-time AR guidance. Compared with image tracking, which primarily relies on planar image markers, three-dimensional object tracking estimates the object's spatial pose based on its geometric contour and natural visual features. This enables the system to register the virtual tooth model with the physical wax block base from multiple viewing angles, thereby more closely matching the actual tooth-carving scenario, in which the operator needs to rotate and observe the wax block during practice. However, this approach usually requires a CAD model or a three-dimensional scanned model of the tracking object, followed by 3D printing and assembly, which increases the complexity of model preparation compared with planar image-marker tracking. Therefore, a practical strategy is to fix the wax block onto a base with distinct geometric structures and visual features. Based on this tracking approach, the system optimizes the tracking strategy by relying on markers assembled into the base to observe the virtual tooth model within a rectangular field of view. During system iteration, a registration threshold adjustment function was added to balance recognition sensitivity and overlay accuracy. A lower threshold improves recognition success but may reduce overlay stability, whereas a higher threshold improves registration accuracy but may increase recognition difficulty. In addition, the tracking interface allows users to switch tracking models and adjust transparency in real time, thereby simplifying operation.

However, this system still has several limitations. First, the current system relies on a computer and an external high-definition camera. The device connection and spatial arrangement are relatively complex, which limits its application in large-class teaching. To reduce cost and improve accessibility, work is underway to adapt the system to mobile devices such as tablets and smartphones. Second, recognition performance is affected by illumination, camera angle, marker occlusion, and base rotation speed. When the marker is occluded or the base is rotated too quickly, the system may fail to recognize it, and the marker must be re-exposed before recognition is restored. Another limitation of this protocol is that several commercial or customized software packages were used for model design, mesh processing, and AR registration. Some functions may be partially replaced by open-source or low-cost alternatives. For example, FreeCAD or Blender may be used for three-dimensional model design and editing, and MeshLab or CloudCompare may be used for mesh processing. In addition, Unity combined with Vuforia Engine may provide a potential development framework for AR based on three-dimensional object recognition, such as model-target-based tracking. However, reproducing the AR workflow used in this protocol would still require additional software development, parameter optimization, and validation. Therefore, although the modeling and mesh-processing steps may be partially reproduced using alternative software, the AR tracking and registration module implemented in this protocol cannot be directly replaced by an existing open-source package without further development and validation.

Finally, the two-dimensional image-plane overlay deviation analysis can only partially reflect the system's overlay accuracy. Since tooth carving is a three-dimensional procedure, the current data cannot fully demonstrate the three-dimensional registration accuracy required for fine tooth-carving training. The effects of marker occlusion, camera angle, illumination, and rotation speed on registration performance were also not separately assessed. Therefore, it remains unclear whether the system provides sufficient accuracy for fine tooth-carving guidance. Finally, this study only completed a single-sample workflow validation, and its ability to improve students’ carving quality and learning outcomes still needs to be confirmed through large-sample controlled studies.

In summary, this study proposes an augmented reality-assisted tooth carving protocol based on three-dimensional object tracking. The single-sample validation results suggest that this workflow can achieve AR overlay between the standard tooth model and the physical wax block operating scene, providing an intuitive three-dimensional visual reference for tooth carving training. Although the current results cannot directly demonstrate that its educational effect is superior to that of traditional methods, this protocol demonstrates the feasibility and value of further research into the application of AR technology to tooth carving training. With advances in mobile devices, three-dimensional recognition algorithms, and artificial intelligence, AR-assisted tooth-carving education may be integrated into dental laboratory teaching in a lower-cost, more convenient, and more scalable manner.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
3D Scanner	SHINING 3D	AutoScanEPCG-BD032G26	Used for three-dimensional scanning of the standard wax tooth.
Camera	Intal RealSense	D415	Used to acquire the real-time visual field for recognition of the three-dimensional object marker.
Dental wax carving instrument			Carve the wax block.
E3D CMF	Hunan Liuwei Jinghang Digital Technology Co., Ltd., Hunan, China.	https://www.e3d-med.com/	AR navigation software.
FDM 3D printer	Bambu Lab	31B8BP631300564	Used to fabricate the wax-block carrier base and three-dimensional object tracking marker.
Geomagic Wrap	3D Systems Corporation	version 2021	Used to process the three-dimensional scan data of the standard wax tooth.
PLA filament	Bambu Lab	10105	Used for FDM 3D printing.
SOLIDWORKS	Dassault Systèmes	version 2025	Used for three-dimensional modeling of the wax-block carrier base and object tracking marker.
Wax block			Tooth carving practice.