Evaluation of an intraoral scan-assisted registration method for guided endodontics: an in vitro study

Abstract

Background

This study was conducted to evaluate the registration accuracy of an intraoral scan-assisted registration (IR) method for guided endodontics.

Methods

Seventy teeth were mounted into five stone dentition models. In terms of the registration method, there were two groups: one utilizing U-shaped tubes for registration (UR group) and the other using an intraoral scanner to aid in registration (IR group). In the UR group, U-shaped tubes were fixed on the model. Preoperative cone-beam computed tomography (CBCT) scans were performed, and fiducial markers on the U-shaped tubes were extracted for registration. In the IR group, fiducial markers were fixed on the crowns after preoperative CBCT scans were performed. The point cloud data were obtained by an intraoral scanner and aligned with the CBCT image to unify the coordinates of the model and markers. The markers were subsequently extracted for registration. The fiducial registration error (FRE) and target registration error (TRE) were calculated to evaluate the registration accuracy. The linear and angular deviations in the accuracy of root canal orifice localization under dynamic navigation were compared and analyzed using the two registration methods. Data were analyzed using t tests and repeated measures analysis of variance (ANOVA) for normally distributed data and the Mann–Whitney U test for nonnormally distributed data.

Results

The TRE of the IR (0.24 ± 0.09 mm) was lower than that of the UR (0.37 ± 0.08 mm), which was reflected mainly in the posterior teeth (P < 0.001). No significant difference in the accuracy of the linear or angular deviation was observed between the two groups (P > 0.05).

Conclusions

Within the limitations of this in vitro study, the IR was an accurate registration method in guided endodontics. However, further clinical validation is needed before this technique can be used in clinical practice.

Background

Canal orifice location in pulp canal obliteration is always time consuming in clinical practice and may cause substantial loss of hard tissue and lead to catastrophic fractures [1, 2]. In recent years, studies have shown that dynamic navigation significantly enhances the accuracy and efficiency of root canal localization following its successful implementation in implantology [3–5]. However, achieving precise localization of root canals is significantly more challenging than implant placement and typically demands a tolerance of less than 1 mm to prevent perforation [6, 7]. Considerable mistakes or even root canal wall perforations have been reported [8–11]. Therefore, additional refinements are necessary to further increase the precision of dynamic navigation in endodontic treatment [3].

The accuracy of dynamic navigation relies primarily on registration precision. The critical step in registration is aligning the cone-beam computed tomography (CBCT) image coordinates—which include the virtual surgical path—with the actual clinical coordinates in a unified coordinate system [12–14]. Marker-based registration is the most commonly used approach in registration techniques; it refers to defining the markers on the CBCT image such that the markers have to be localized with the probe of the navigation system in physical space for coordination transformation [12, 15–17]. Fiducial registration error (FRE) and target registration error (TRE) are two key metrics for evaluating registration accuracy [12–14]. The FRE is defined as the distance between corresponding fiducial markers after registration and is an indicator of the quality of registration given by the computer [13]. The TRE is the distance between each corresponding point other than the fiducial marker and represents the practical application accuracy of dynamic navigation [13].

The U-shaped tube registration (UR) method is a widely adopted marker-based registration technique because of its reliability and accuracy [15–17]. Prior to the surgical procedure, a U-shaped tube equipped with preestablished markers is affixed firmly to the designated surgical site using silicone rubber. A CBCT scan is subsequently performed to obtain a virtual image that incorporates the marker information. Prior to initiating the navigation process, the U-shaped tube should be carefully repositioned to ensure that its markers are accurately aligned with the corresponding markers in the CBCT image. However, this study has certain limitations. First, the silicone rubber used to secure the U-shaped tube may deform during storage, which could result in registration errors and consequently compromise the accuracy of the navigation system [15, 16]. Second, if a patient has already undergone CBCT imaging for diagnostic and examination purposes prior to surgery, requiring the patient to wear a U-shaped tube during subsequent CBCT scans results in increased radiation exposure [3, 15, 16].

High-precision images can be obtained via intraoral scanning [18, 19]. Some scholars have utilized this technology to extract dental scanning data and CBCT images for image registration in guided implant surgery, thereby increasing the accuracy of registration [20, 21]. However, unlike the registration target in implant surgery, which is situated on the surface of the jawbone below the occlusal plane, the registration target for endodontic treatment is located on the tooth. Moreover, the distinct morphological and structural characteristics of anterior and posterior teeth may influence registration and drilling accuracy. Therefore, this study presents the use of intraoral scanning to assist with registration in guided endodontic treatment, addressing the limitations of the U-shaped tube method and potentially enhancing registration accuracy. The markers are attached to the tooth crowns within the target surgical area, and intraoral scanning is performed to capture dental data that include these markers, enabling subsequent image registration with the preoperative CBCT scan. The marker information is subsequently incorporated into the CBCT imaging process. Analogous to U-shaped tube registration, markers located on the teeth are identified and matched with their corresponding virtual markers in the CBCT images for the subsequent marker-based registration process. This method, referred to as intraoral scan-assisted registration (IR), reduces errors associated with the secondary positioning of the U-shaped tube. Additionally, if a patient has already undergone a CBCT scan, the need for additional CBCT imaging is eliminated, thereby reducing radiation exposure. It remains uncertain whether this IR technique is applicable to guided endodontics and how its accuracy is comparable to that of the conventional UR approach.

Therefore, the aim of this study was to compare the registration accuracy and canal localization accuracy of IR and UR in extracted human teeth to explore the feasibility of its application in guided endodontic treatment. The null hypothesis of this study was that there would be no difference in registration accuracy between IR and UR.

Methods

Model Preparation

The workflow of this research is shown in Fig. 1. In accordance with the registration method, the experiments were divided into two groups: the UR group and the IR group.

Fig. 1.

Flow chart

The sample size was calculated using G Power (version 3.1.9.7; Heinrich-Heine-Universität Düsseldorf, Germany). To compare the registration accuracy between the two groups, the two registration methods were applied to the same model. A sample size calculation suitable for a paired-samples t test was performed, using an alpha level of 0.05 and a statistical power of 0.95. The results of a pilot study revealed a mean TRE of 0.36 mm for the UR group and 0.25 mm for the IR group; n = 4 per group was needed. To compare the deviations in navigation between the two groups, a sample size calculation suitable for an independent samples t test was performed, with an alpha value of 0.05 and a statistical power of 0.8. In a pilot study, the mean linear deviation at the apical point was 0.63 mm for the UR group and 0.47 mm for the IR group; n = 22 per group was needed.

Seventy disarticulated permanent teeth were collected for periodontal reasons from the First Clinical Division of Peking University School of Stomatology. The inclusion criteria included teeth with well-defined crown anatomy and clearly visible root canals on CBCT images and the absence of caries, significant wear, or restorations. One mandibular stone model, containing six mandibular incisors, four mandibular premolars, and four mandibular molars, was constructed to evaluate registration accuracy. Two complete dentition models of the maxilla and mandible—comprising 24 incisors, 16 premolars, and 8 molars—were constructed to assess the precision of canal localization using the two registration methods. All the teeth were positioned in anatomically correct locations. The institutional review boards of the Peking University School and Hospital of Stomatology approved their use (Approval Number: PKUSSIRB-2021106904).

Registration

One mandibular model was applied to evaluate registration accuracy. Within each group, the registration areas were divided into anterior and posterior tooth regions. A pit measuring 1 mm in diameter and 1 mm in depth was created on the lingual fossa of the left mandibular first incisor and on the occlusal surface of the left mandibular first molar, serving as target points for TRE measurement for each region. These target points were utilized to simulate the access cavity entry points for each tooth.

For the UR group, U-shaped tubes (DCarer, Suzhou, China) designed for the anterior and posterior teeth regions were placed in the corresponding positions of the model with silicone rubber (DMG Dental Products, Hamburg, Germany) (Fig. 2a and b). Preoperative small field-of-view CBCT (CS9300; Carestream Health, Rochester, NY, USA) was performed in the anterior and posterior teeth regions with a 90 μm slice thickness. The digital imaging and communications in medicine (DICOM) data were uploaded into Mimics Medical software 21.0 (Materialise, Leuven, Belgium). The model was segmented from the CBCT data and reconstructed in 3D for registration purposes. According to the instructions, six center points of pits on the U-shaped tube were defined as fiducial markers in the 3D CBCT reconstruction for the anterior teeth and posterior teeth regions. The fiducial markers were arranged by considering the following key guidance for fiducials [14]: (1) avoid nearly linear placement; and (2) ensure that the center point of the fiducial markers is as close as possible to the desired location (Fig. 2c and d).

Fig. 2.

3D reconstruction of the UR group. a & b U-shaped tubes were fixed to the anterior and posterior teeth regions with silicone rubber. c & d the U-shaped tubes of the anterior and posterior regions were segmented from the CBCT data and reconstructed in 3D for registration purposes. Six fiducial markers were selected for each region. Red points indicate fiducial markers selected for registration

An emery bur (TR11; Mani, Tochigi, Japan) was fixed to a high-speed handpiece (NSK, Nakanishi, Tochigi, Japan) with an infrared emitter (DCarer). The handpiece was calibrated within the visual range of the dynamic navigation system (Polaris; Northern Digital, Ontario, Canada). Afterward, we identified the six fiducial markers corresponding to the CBCT image data and used the handpiece to extract the physical coordinates of the anterior and posterior teeth separately. Each procedure was repeated ten times. The observer practiced the registration procedure 20 times before the study started. The coordinates of the fiducial markers in the CBCT data and in physical space were stored in the TXT format for further analysis.

The FRE and TRE were subsequently computed to evaluate the registration accuracy. The data were transferred into MATLAB R2021a (MathWorks, USA) for accuracy analysis. Singular value decomposition (SVD) was applied to the coordinates of the markers for registration. The FRE was calculated as the root mean square distance between homologous fiducial markers (Fig. 3a). To calculate the TRE, the target points preset on the model were identified in the 3D CBCT reconstruction and in the physical space using the calibrated handpiece. The TRE was calculated as the spatial Euclidean distance between a target point after coordinate transformation and the bur point approaching the homolog target point in the physical space (Fig. 3b). For each target point, the TRE was calculated over ten measurements after each registration procedure.

Fig. 3.

Schematic diagram of FRE and TRE calculation. a The FRE is calculated as the root mean square distance between homologous fiducial markers after registration. The blue solid circles denote fiducial markers in the physical space extracted by the handpiece under navigation. The black hollow circles represent the measured positions of corresponding fiducial markers after registration. b The TRE is the Euclidean distance between the target points after registration. The green solid circle and green hollow circle denote the target point in physical space obtained by the handpiece under navigation and the corresponding target point after coordinate transformation within the registered coordinate system. The red arrows represent the error value between two points

For the IR group, fiducial markers were attached to the crowns after the preoperative CBCT scan. Dental data, including these markers, were obtained by an intraoral scan, and image registration was performed with preoperative medical CBCT images. Afterward, the markers were selected on the CBCT image, and the corresponding markers were located in the physical space using a navigation probe for registration to unify the coordinate systems of both. The specific details are as follows.

CBCT scans were performed with the same exposure parameters as those of the UR group for the anterior teeth region and posterior teeth region. No additional device was mounted during the CBCT imaging. The DICOM data were imported into Mimics Medical software 21.0 (Materialise), and 3D reconstructions were performed (Fig. 4a and b).

Fig. 4.

Image registration between CBCT images and intraoral scan images. a & b Small field-of-view CBCT scans of the anterior and posterior teeth regions were acquired and three-dimensionally reconstructed. c & d Intraoral scan images of the anterior and posterior teeth regions were obtained after fiducial markers were attached to the crowns. e & f ICP registrations were performed to align the CBCT image with the intraoral scan images of both the anterior and posterior teeth regions. Red points indicate fiducial markers selected for the subsequent marker-based registration

Customized composite resin blocks (Filtek Z350XT, 3 M ESPE, Saint Paul, MN, USA), featuring a central pit measuring 1 mm in diameter and 1 mm in depth, were bonded to the crown surfaces as fiducial markers. Six such markers were strategically placed in both the anterior and posterior teeth regions. The fiducial markers were also arranged in accordance with the key guidelines for fiducial placement [14]. In the anterior region, markers were placed on the incisal ridge and cingulum of the left canine, left incisor, and right canine. In the posterior region, markers were placed on the buccal and lingual cusps of the left premolar, the buccal and lingual grooves of the left first molar, and the buccal and lingual cusps of the left second molar.

An intraoral scanner (DL100; Launca Medical Device Technology, Guangzhou, China) was used to scan the model to obtain standard tessellation language (STL) data of the model with fiducial markers (Fig. 4c and d). The data were imported into 3-Matic Research software 13.0 (Materialise, Leuven, Belgium) for registration with 3D reconstruction of the preoperative CBCT data. After the initial rough alignment, iterative closest point (ICP) registration was performed to align the 3D point clouds of the two images and transform the geometry of the fiducial markers on the model to correspond with the CBCT space (Fig. 4e and f). As a result, the base coordinate system of the fiducial markers and model was the same as that of the CBCT space. After this procedure, the image registration error (IRE) presented by the software was documented. Afterward, the coordinates of each fiducial marker were determined manually on the aligned image in Mimics Medical software 21.0 (Materialise, Leuven, Belgium) and extracted by the calibrated handpiece in physical space. Then, SVD was also applied to the coordinates of the markers for registration. The aforementioned steps were repeated a total of ten times.

After each registration procedure, the IRE was documented, and the FRE and TRE values were calculated. The calculation method was the same as that in the UR group. A single operator was used to ensure consistency in technique and reduce variability.

Real-time navigation

The remaining two sets of complete dentition models made from disarticulated teeth were randomly divided into two groups. Each set of models had 28 teeth. Small field-of-view CBCT scans (CS9300; Carestream Health, Rochester, NY, USA) were performed in the anterior and posterior tooth regions with a 90 μm slice thickness. DICOM data were uploaded into Mimics Medical software 21.0 (Materialise). One straight-line access cavity was planned for each tooth, extending from the occlusal surface to the orifice of a single canal. A total of 28 drilling paths were designed for each group, which exceeds the minimum sample size requirement.

Two registration methods were applied within each group. Access cavities were prepared using calibrated high-speed handpiece burs (TR11; Mani) with real-time navigation guidance (Polaris) under continuous water cooling. Since there is a learning curve for dynamic navigation, the operator practiced dynamic navigation on 20 teeth before drilling. After access cavity preparation, the canals were located using a #10 C-file (C-Pilot; VDW GmbH, Munich, Germany).

Postoperative measurement

Postoperative CBCT scans with the same exposure parameters as those used for the preoperative CBCT images for each model were obtained. The DICOM data were segmented, and 3D reconstruction was performed using Mimics Medical software 21.0 (Materialise). The entry point and apical point of the central axis of the actual drilling path were identified in axial, sagittal, and coronal views to establish the precise trajectory of the drilling path. The 3D-reconstructed data were matched with the preoperative 3D-reconstructed CBCT data via 3-Matic Research software (Materialise). The deviations at both the entry point and the apical point, as well as the angular deviations between the actual and planned drilling paths, were measured using 3-Matic Research software (Materialise) (Fig. 5).

Fig. 5.

Illustration of parameters indicating deviations between the actual drilling path (the blue cylinder) and the planned drilling path (the orange cylinder). ① indicates linear deviation at the entry point; ② indicates linear deviation at the apical point; and ③ indicates angular deviation

Statistical methods

The data analysis was conducted by an independent individual who did not participate in the experiment and was blinded to the group assignments. Analysis was performed with SPSS 24.0 (IBM, Armonk, NY, USA). The data are expressed as the means ± standard deviations. Distribution normality was assessed using the Shapiro–Wilk test. If the data satisfied the assumptions of normality and homogeneity of variance, a paired-samples t test was employed to compare registration accuracy between the IR group and the UR group, whereas an independent samples t test was conducted to assess differences in root canal localization accuracy between the two groups; otherwise, the Mann‒Whitney U test was employed. Repeated measures analysis of variance (ANOVA) was used to evaluate differences among the three variables—IREs, FREs, and TREs—in the IR group. The significance level was set at P = 0.05.

Results

The registration accuracy of the IR group is shown in Table 1. The data collected in this study satisfied the assumptions required for conducting a repeated measures analysis of variance. The differences among the three registration error values within the IR group were statistically significant (P < 0.05). Postevent inspection revealed that the IREs were significantly lower than both the FREs and the TREs (P < 0.05).

Table 1.

Registration accuracy in the IR group (mm)

	IRE		FRE		TRE
	Mean ± SD	95% CI	Mean ± SD	95% CI	Mean ± SD	95% CI
Anterior teeth region	0.05 ± 0.00	0.05, 0.05	0.21 ± 0.03	0.19, 0.22	0.32 ± 0.04	0.30, 0.34
Posterior teeth region	0.05 ± 0.00	0.05, 0.06	0.22 ± 0.03	0.21, 0.24	0.17 ± 0.04	0.15, 0.19

95% CI  95% confidence interval

A comparison of the registration accuracies of the two registration methods is shown in Fig. 6. The TRE of the IR group (0.24 ± 0.09 mm) was significantly lower than that of the UR group (0.37 ± 0.08 mm) (P < 0.001). The TRE of the posterior teeth in the IR group (0.17 ± 0.04 mm) was significantly lower than that in the UR group (0.40 ± 0.06 mm) (P < 0.001), whereas the TRE of the anterior teeth in the IR group (0.32 ± 0.04 mm) was not significantly different from that in the UR group (0.35 ± 0.10 mm) (P = 0.470). The FREs of both the anterior and the posterior teeth did not significantly differ between the two registration methods (P > 0.05).

Fig. 6.

Comparison of the FREs and TREs. a overall deviation of the TREs between the UR and the IR group; (b) deviation of the TREs in the anterior teeth region between the UR and the IR group; (c) deviation of the TREs in the posterior teeth region between the UR and the IR group; (d) overall deviation of the FREs between the UR and the IR group; (e) deviation of the FREs in the anterior teeth region between the UR and the IR group; (f) deviation of the FREs in the posterior teeth region between the UR group and the IR group. A horizontal bar linking columns indicates a significant difference (P < 0.001)

All the root canals were successfully located. No significant difference was found between the two registration methods in terms of linear or angular deviation (P > 0.05) (Table 2).

Table 2.

Metric deviations between the planned and actual drilling paths between the two groups

	Deviation at the entry point (mm)		Deviation at the apical point (mm)		Angular deviation (°)
	Mean ± SD	95% CI	Mean ± SD	95% CI	Mean ± SD	95% CI
UR	0.47 ± 0.25	0.37, 0.56	0.52 ± 0.22	0.43, 0.60	6.02 ± 3.35	4.72, 7.32
IR	0.51 ± 0.21	0.42, 0.59	0.42 ± 0.21	0.34, 0.50	6.58 ± 3.69	5.15, 8.01
P	0.518		0.106		0.589

Discussion

The results of this study demonstrated that the registration accuracy of the IR is superior to that of the UR, thereby rejecting the null hypothesis.

The IRE of the IR is remarkably low. This method involves scanning the dentition containing markers via a high-precision intraoral scanning device and then converting the information of the markers into CBCT images by graphically fitting the dentition images to the CBCT images using the ICP algorithm. The results revealed that the image registration accuracy was high, with an average error of only 0.05 mm. ICP is among the most commonly used point cloud image registration methods and can reduce the minimum point-to-point distance between two images and minimize the gap between them [22]. The accuracy of the ICP registration is affected by the minimum resolution of the image [22]. The average scanning accuracy of intraoral scanning is approximately 20 μm. The small-field CBCT utilized in this study possesses an image resolution of 90 μm, thereby enabling the acquisition of high-precision 3D reconstructed images through meticulous image segmentation. The IRE of the ICP point cloud fitting of the two images is much lower than that of the fitting of the medium-field CBCT image and the intraoral scanning image by Kim et al. [23] (0.234 mm). The FREs and TREs of the IR group were significantly greater than those of the IRE group. This might be because markers were adopted for the subsequent registration process in the IR group, and the manual selection of markers generated error. Therefore, in the future, if a dynamic navigation system can automatically register the optically scanned target area, greatly improved registration accuracy is expected.

The results of this study revealed that the FRE value of the IR was comparable to the error value of the UR. Marker-based registration aligns the entire anatomical structure of the surgical area by selecting virtual markers in the CBCT image and fitting them to the corresponding markers in the physical space [12–14]. The FRE represents the discrepancy between the selected virtual markers and their corresponding actual markers after registration, with the magnitude of this discrepancy being dependent on the accuracy of point acquisition [12–14]. This is a prerequisite for ensuring the accuracy of the actual registration process. The high image resolution of the CBCT image with a small field of view enables precise positioning of the virtual markers in both groups. Additionally, the actual markers are preset on the surgical area in both registration methods, ensuring accurate and stable selection of the center of the corresponding markers during the actual operation. The low FREs in this study suggest a high level of registration accuracy for both registration methods, thereby validating their effectiveness in achieving precise registration.

The TRE is used to test the registration accuracy of the marker-based registration and serves as an indicator of the ultimate precision achieved in registration. Lower TREs were observed in the posterior teeth when the IR was applied than when the UR was applied. The TRE is influenced not only by the accuracy of marker extraction but also by the relative positional relationship between the markers and the target points [12–14, 24]. West et al. [13, 14] demonstrated that the distribution of markers close to target points could reduce the TRE. These results might have occurred because the markers located on the crown were closer to the target points than the markers on the U-shaped tube were. However, the two registration methods resulted in the same TRE in the anterior teeth; this might be because although the markers in the IR group were closer to the target points than those in the UR group were, the localization error introduced when the markers and target points on the inclined surfaces of the anterior teeth were identified could have reduced the TRE in the IR group.

The average distance deviation value in navigation was approximately lower than 0.6 mm under the IR. Because the dynamic technique involves a learning curve, substantial training is needed before it is applied [10, 25, 26]. The operator conducted 20 training sessions on extracted teeth before formal drilling to avoid the influence of human factors [25, 26]. The accuracy of canal location was not significantly different between the two registration methods; this may be attributed to the fact that the drilling accuracy under navigation is influenced not only by the registration accuracy but also by human operational errors. Nevertheless, the IR method eliminates potential errors caused by repositioning of the U-shaped tube and the need for an additional CBCT scan. Furthermore, this method is unaffected by soft tissue and allows for the arrangement of markers according to different target points, which is consistent with the advantages of the gold standard bone-screw registration method for implantation [12, 17].

This study is subject to several limitations. First, this study used an in vitro setup. The experimental conditions used in in vivo and in vitro studies can significantly influence outcomes [27]. In real clinical scenarios, slight movement of the patient’s head and jaw may affect the accuracy of marker collection, thereby influencing registration accuracy. In addition, during operation, the navigation system’s inherent tracking accuracy for motion and the operator’s hand stability can influence the final positioning accuracy [12]. Second, the jaw models might fail to replicate actual tooth wear, the presence of metallic prostheses and other realistic dental conditions. Therefore, further extensive clinical research is necessary to confirm these results.

Conclusions

The IR can achieve a precise registration accuracy with a mean TRE value lower than 0.4 mm, which is lower than that of the UR in the posterior teeth region. The mean accuracy of root canal localization using IR was less than 0.5 mm, thereby reducing the risk of perforation. In addition, IR can avoid the errors introduced by UR and avoid additional CBCT scans. Within the limitations of this in vitro study, the IR showed sufficiently high registration accuracy in guided endodontics and holds promise for clinical applications. However, further clinical validation is needed before this technique can be effectively implemented in clinical practice.

Abbreviations

CBCT

Cone beam computed tomography

UR

U-shaped tube registration

IR

Intraoral scan-assisted registration

FRE

Fiducial registration error

TRE

Target registration error

DICOM

Digital imaging and communications in medicine

SVD

Singular value decomposition

STL

Standard tessellation language

ICP

Iterative closest point

IRE

Image registration error

95% CI

95% confidence interval

Authors’ contributions

C.L., L.M., X.B. and Y.D. conceptualized the overall strategy. C.L. and L.M. contributed to planning and execution. C.L. performed the statistical analyses, including the figures. C.L. and L.M. contributed to the investigation. C.L. and L.M. contributed to the methodology. C.L. and X.B. wrote and prepared the original draft. C.L., X.B. and Y.D. contributed to review and editing. X.B. and Y.D. provided supervision. All the authors have read and agreed to the published version of the manuscript.

Funding

This work was funded by the Peking University Hospital of Stomatology (Grant PKUSSNKP-2019 and Grant YT-2021-7); the Natural Science Foundation of Beijing Municipality (L252201, L252126); and the National Key Research and Development Program of China (2024YFC2418101).

Data availability

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All procedures performed in studies involving participants (whose teeth were used) were in accordance with the ethical standards of the Ethics Committee of Peking University School and Hospital of Stomatology and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This study was approved by the Ethics Committee of Peking University School and Hospital of Stomatology (Approval Number: PKUSSIRB-2021106904). The requirement for consent to participate was waived.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.