Computer-aided design and manufacturing of orthodontic retainers from digitally debonded models can be used to facilitate same-day delivery. The purpose of this prospective clinical study was to validate a novel technique for virtual bracket removal (VBR) in-office, comparing the accuracy with 2 orthodontic laboratories that use VBR for retainer fabrication in the digital workflow.
The sample consisted of 40 intraoral scans of 20 patients. Four groups were compared. The scans without brackets were used as a control group. VBR was performed by 3 groups: In-office VBR (Software Meshmixer, version 3.5.474; Autodesk, San Rafael, Calif), Orthodent Laboratory (ODL; Buffalo, NY), and New England Orthodontic Laboratory (NEOLab; Andover, Mass). The virtually debonded models were superimposed onto the control models using surface-based registration. Regional 3-dimensional Euclidean distances between surface points of superimposed models were calculated for comparative analysis of surface changes after VBR using Vector Analysis Module (Canfield Scientific, Fairfield, NJ) software.
The accuracy of VBR using the Meshmixer did not differ significantly from the VBR protocols used by the 2 laboratories. However, there was a statistically significant difference between the 2 laboratories, with ODL showing lower accuracy than NEOLab. Although some differences were statistically significant, they were very small and not considered clinically relevant. There was also a statistically significant difference between the 3 tooth segments (incisors, canines/premolars, and first molars), with VBR of the first molars and second premolars showing the least accuracy.
The VBR techniques using the in-office Meshmixer, ODL, and NEOLab were considered accurate enough for the clinical use of orthodontic retainers fabricated from printed models.
A novel technique for virtual bracket removal (VBR) for modern orthodontic retainers.
Accuracy of VBR for orthodontic retainers from 3-dimensional printed models.
Digitally removed brackets for orthodontic retainers in the digital workflow.
VBR with Meshmixer is accurate for retainer fabrication.
One of the greatest orthodontic challenges is maintaining tooth position after debonding, and thus, ensuring timely manufacturing of retainers is key to the success and longevity of orthodontic treatment. Orthodontic retainers should be placed immediately after the removal of the appliances because some relapse may occur in a few hours. , Traditionally, retainer fabrication workflow has involved taking an alginate impression before bracket removal or after bracket removal, pouring the impression up in stone, physically carving the brackets off if the impression was taken before bracket removal, and fabricating the retainer on the stone model. Problems during the stone model fabrication may occur, requiring the patient to come back to the office for a new impression. With the introduction of intraoral scanners and software for computer-aided design (CAD) and computer-aided manufacturing, orthodontic appliance fabrication techniques have evolved and have become digital. , Advantages of 3-dimensional (3D) digital scanning include simplicity, accuracy, longevity, reduced patient discomfort, elimination of impression material in inventory, reduced storage issues, and minimization of cross-contamination. Moreover, the intraoral scan can be 3D printed in resin and used for appliance fabrication. With 3D printing gaining traction in the orthodontic community, many private practices have been investing in in-office digital laboratories for 3D-printing models and fabrication of appliances to increase efficiency and reduce the number of appointments for patients. Using the digital workflow, high-quality retainers can be fabricated from 3D printed models.
In the digital workflow, the retainer fabrication involves the acquisition of the patient’s intraoral scan, postprocessing of the digital models in stereolithography (STL) file format, virtual bracket removal (VBR) procedure in CAD software, model 3D-printing, and fabrication of the retainer before the debonding appointment. The first step in this workflow is to acquire an accurate intraoral scan, which is critical. There are many intraoral scanners in the market that possess trueness and precision sufficient for orthodontic applications. Among them, TRIOS 3 (3Shape, Copenhagen, Denmark) is a popular intraoral scanner that has shown good trueness and precision even when scanning arches with bonded buccal brackets. Studies have shown that digital impressions and models accomplish equal or higher precision than some conventional impression materials and stone models. After the acquisition of the intraoral scan, digital model postprocessing is necessary to remove artifacts such as noise, outliers, holes, or ghost geometry, , preparing the models for the next step: VBR, which is a new procedure in orthodontics in which the brackets are digitally selected and removed from the tooth surface to produce a digital model without brackets. VBR can be performed using many different CAD software programs, such as Meshmixer (Autodesk, San Rafael, Calif) and OrthoAnalyzer (3Shape). VBR may take 4-5 minutes per arch, depending on the software used and the operator’s skill. Once VBR is performed, the digital model can be 3D printed to serve as a physical model for retainer fabrication. VBR before the debonding appointment has the important advantage of same-day delivery of a well-fitting fixed or removable retainer, with the added advantage of eliminating an office visit.
VBR can be performed either in-office or by an orthodontic laboratory that provides this new digital service in lieu of physical carving of brackets from traditional plaster models. Because VBR is a novel technique, there are no studies that have tested this procedure to produce accurate 3D-printed models for retainer fabrication. Thus, this prospective clinical study aimed to validate a novel technique for VBR in-office, comparing the accuracy with 2 orthodontic laboratories that use VBR for retainers fabrication in the digital workflow.
The hypothesis is VBR can be performed accurately enough to be used for orthodontic retainers, independently of the software program or laboratory.
Material and methods
The University Park Institutional Review Board of the University of Southern California (USC) approved this study. Informed consent and assent were obtained from the legal guardians and patients, respectively. The inclusion criteria for this prospective clinical study were as follows: (1) patients starting or finishing orthodontic treatment, allowing for intraoral scans done on the same day as bonding or debonding, (2) full fixed labial appliances bonded at least from maxillary left first molar to maxillary right first molar, and (3) at least one tooth per segment (incisors, canine/premolars, and first molar). In this study, 2 patients were excluded because of poor scan quality and the presence of a band on a first molar. The sample consisted of 40 maxillary dentition intraoral scans of 20 patients of the USC Advanced Orthodontic Clinic. Two maxillary intraoral scans of each patient, one with brackets and one without brackets, were acquired at the same appointment during either the beginning (before and after bonding, 2 of 20 patients) or at the completion of orthodontic treatment (before and after debonding, 18 of 20 patients) using the TRIOS 3 intraoral scanner. The scanner’s software postprocessed the intraoral scans, exporting them as digital models in the STL format. Four groups were compared (all subject participants were included in each group). Group 1 was the control group that consisted of intraoral digital models without brackets (postdebonding scans/prebonding scans). Group 2 was post-VBR digital models from the in-office Meshmixer VBR protocol. Group 3 was post-VBR digital models from Orthodent Laboratory (ODL; Buffalo, NY), and group 4 was post-VBR digital models from New England Orthodontic Laboratory (NEOLab; Andover, Mass) ( Fig 1 ).
The Meshmixer VBR protocol used in this study was developed and validated in vitro in a previous preliminary typodont study. Meshmixer is freeware software that can be downloaded at www.meshmixer.com . After importing the STL files into the Meshmixer software, the first step was the model preparation, which includes digitally removing scan artifacts connected to brackets before VBR. Per the VBR protocol, surface lasso selection mode allows surface faces to be selected without “painting.” These selection boundaries can be refined using the smooth boundary tool. Once the boundary around the bracket is smoothed, the erase and fill tool is selected to virtually erase the bracket. Figure 2 details the protocol for VBR using Meshmixer, which was performed from maxillary right first molar to maxillary left first molar.
There are a few outside orthodontic laboratories that offer digital bracket removal. Two of these laboratories were selected on the basis of their advertised capability to digitally remove brackets: ODL and NEOLab. Each subject participant’s digital model was coded with a number as not to reveal the subject’s name (eg, VBR #2). All digital models of the bracketed maxillary arch for each subject participant were sent to both laboratories. These laboratories performed digital bracket removal using OrthoAnalyzer software with their own VBR protocols and attached a digital model in STL format for each patient after bracket removal. To avoid the risk of bias, neither laboratory was aware of the study.
For 3D evaluation of VBR accuracy, the digital models of all groups were imported into the 3-matic 3D modeling software (Materialise, Leuven, Belgium) for 3D superimposition. Although STL models do not contain any volumetric data, their triangulated surface data can be used for the superimposition of 3D surface data of digital models. The scans of groups 2-4 were superimposed onto the group 1 control models using the surface-based registration technique, which provides the best fit between the models ( Fig 3 ). The superimposition accuracy on the areas not affected by the VBR was evaluated by the iterative closest point algorithm and color-coded maps (± 300 μm visualization range). To complement the visual inspection, the measure analysis locally tool was used to quantify the superimposition error in the stable areas ( green color ) where surface changes were not expected ( Fig 3 , D ). If the registration error ranged from 0 to 0.05 mm, the model’s superimposition accuracy was confirmed. Then, the digital models were exported as STL files and transferred to the Vector Analysis Module (VAM; Canfield Scientific, Fairfield, NJ) for 3D assessment of the VBR accuracy.
The entire VBR procedure was performed once by each group. All the measurements for VBR accuracy assessment (40 maxillary intraoral scans) were performed by 2 separate investigators using VAM to ensure interexaminer reliability. The investigators were third-year orthodontic residents previously trained by 1 experienced orthodontist (AW) on how to use the software for VBR evaluation. Each investigator repeated all of the measurements twice (40 maxillary intraoral scans) after a 2-week interval to ensure intraexaminer reliability. Four groups of maxillary arch digital models were thus identified for the measurements: (1) a control group of clinically debonded or prebonding, (2) VBR performed in-office by Meshmixer protocol, (3) VBR performed by ODL, and (4) VBR performed by NEOLab. Regional 3D Euclidean distances between surface points of the superimposed control and debonded models were measured using VAM software. The superimposed control and virtually debonded models were measured on the labial surface using an iterative closest point algorithm for comparative analysis of surface changes after VBR. The color surface by distance tool (± 300 μm visualization range) was used to display and to indicate the areas with surface changes to be included in the measurement. For each tooth, the paint area selection tool was used to select the area where the bracket was virtually removed, previously indicated by the color maps ( Fig 4 ). A regional color-coded map (± 300 μm visualization range) of the selected area illustrated the linear surface changes after VBR. VAM automatically calculated the minimum, maximum, root mean square (RMS), and mean values with a standard deviation for each selected area. Comparisons were made between the 3 VBR techniques (Meshmixer, ODL, and NEOLab), the tooth segments, and individual teeth. An overview of the whole workflow for the VBR process and 3D evaluation is shown in Figure 5 .
The 3 models that underwent VBR in each sample were randomized and coded by 1 of the authors to ensure blinding of the examiners performing the measurements. The 3 VBR models in each sample were given the codes U, X, or Y, each corresponding to a specific laboratory that was unknown by the examiner performing the measurements. The coding system was revealed only after the statistical analysis was complete.
The RMS values, which best represented the overall magnitudes of surface change irrespective of the direction of change of the 3 techniques, were compared. Descriptive statistical analysis was performed with SPSS (version 20.0; IBM, Armonk, NY) for linear surface changes because of VBR. Interexaminer (Cronbach alpha) and intraexaminer (Cronbach alpha) reliability were determined. The Shapiro-Wilk test was used to evaluate the normality of the data. Multiple linear regression analysis with tooth/segment and laboratory as independent variables was first attempted. Because the laboratory was identified to contribute insignificantly to the total variance, 3 separate 1-way analysis of variance (ANOVA) tests were used to detect the potential differences between 3 VBR protocols/laboratories, separate teeth, and tooth segments (incisors, canines/premolars, and first molars). Statistical differences between the test groups were further analyzed with Scheffé post-hoc test (⍺ = 0.05).
Interexaminer and intraexaminer reliability were determined to be high (>0.9). Because there were no significant interexaminer and intraexaminer differences in the VBR measurement reliability, the RMS values from both operators and measurement trials were averaged by tooth for each VBR group. One-way ANOVA used averaged RMS surface changes by tooth, from central incisor to first molar (1-6), and significance was determined at P <0.05. RMS surface changes ranged from 0.10 mm to 0.30 mm, with first molars exhibiting the greatest amount of surface change and central incisors exhibiting the least amount of surface change resulting from VBR ( Fig 6 ). The first molars and the second premolar showed the greatest distribution of surface change, indicating the largest error in VBR. Post-hoc analysis with Scheffé test ( Table I ) showed a pairwise statistically significant difference ( P <0.05) in averaged RMS values between the second premolar and all other teeth in the arch as well as the first molar and all other teeth in the arch.