If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding authors: Dhelal Al-Rudainy: University of Baghdad Bab Al-Moadham Campus, College of Dentistry, Iraq. Liu Yang: The Eighth Affiliated Hospital of Sun Yat-Sen University, Shenzhen, Guangdong 518003, China.
Corresponding authors: Dhelal Al-Rudainy: University of Baghdad Bab Al-Moadham Campus, College of Dentistry, Iraq. Liu Yang: The Eighth Affiliated Hospital of Sun Yat-Sen University, Shenzhen, Guangdong 518003, China.
Three-dimensional (3D) modeling of orthodontic dental casts can be accomplished by smartphone-based stereophotogrammetry.
•
The accuracy of virtual orthodontic dental casts produced by smartphone-based stereophotogrammetry was about 0.34 mm, and the error of repeated 3D models using this method was 0.03 mm.
•
Stereophotogrammetry using smartphone devices is a simple and low-cost technology for the 3D modeling of orthodontic dental casts.
•
This method is a helpful low-cost tool for the digital archiving of dental casts, eliminating physical storage shortcomings, such as broken models and space requirements.
ABSTRACT
Background
The development of intraoral scanning technology has effectively enhanced the digital documentation of orthodontic dental casts. Albeit, the expense of this technology is the main limitation.
The purpose of the present study was to assess the validity and reliability of virtual three-dimensional (3D) models of orthodontic dental casts, which were constructed using smartphone-based 3D photogrammetry.
Methods
A smartphone was used to capture a set of two-dimensional images for 30 orthodontic dental casts. The captured images were processed to construct 3D virtual images using Agisoft and 3DF Zephyr software programs. To evaluate the accuracy of the virtual 3D models obtained by the two software programs, the virtual 3D models were compared with cone-beam computed tomography scans of the 30 dental casts. Colored maps were used to express the absolute distances between the points of each compared two surfaces; then, the means of the 100%, 95th, and 90th of the absolute distances were calculated. A Wilcoxon signed-rank test was applied to detect any significant differences.
Results
The differences between the constructed 3D images and the cone-beam computed tomography scans were not statistically significant and were accepted clinically. The deviations were mostly in the interproximal areas and in the occlusal details (sharp cusps and deep pits and fissures).
Conclusions
This study found that smartphone-based stereophotogrammetry is an accurate and reliable method for 3D modeling of orthodontic dental casts, with errors less than the accepted clinically detectable error of 0.5 mm. Smartphone photogrammetry succeeded in presenting occlusal details, but it was difficult to accurately reproduce interproximal areas.
In orthodontics, dental casts are valuable tools for precise diagnosis, treatment planning, and evaluation of treatment outcomes. The British Orthodontic Society has recommended retaining orthodontic study models for 11 years or until young patients reach 25 years old [
]. Digital intraoral scanning devices have provided three-dimensional (3D) virtual models of dentition that offer numerous applications in dentistry generally and in orthodontics specifically [
. This innovative technology helps orthodontists in many aspects, such as digital analysis of dental arches, occlusion evaluations, and digital storage of study models. However, this technology is not readily available in every dental clinic because of its high cost; the average expenditures are between £13,000 and £31,000 [
]. It has been concluded that photogrammetry is applicable for the digitization of dental models, and further studies have been suggested to evaluate the clinical application of its methods.
Smartphone technology has become part of our daily lives and requires no professionality or experience. It has been presented in orthodontic literature for different purposes, such as digital cephalometric analysis [
]. Virtual facial models were created by capturing two-dimensional (2D) images of the face using a smartphone, and, with the aid of software, 3D models of the face were generated [
. The technique was noninvasive, and it was considered a low-cost alternative to stereophotogrammetry and 3D scanners. The application of this technology has not been limited to 3D facial construction; it was used by Barbero-García et al. (2019) [
] for the production of virtual cranial vaults for 10 children with cranial deformities. Using a smartphone camera, short videos were recorded, and then, a series of 2D images were extracted from the videos to create 3D models of the cranial vaults using PhotoScan (Agisoft) software. The accuracy of the obtained 3D meshes was comparable to medical diagnostic imaging of cranial vaults (magnetic resonance imaging and computed tomography scans) [
To the best of our knowledge, in literature, smartphone-based photogrammetry has not yet been used for the creation of 3D virtual models of orthodontic dental casts. This technology can offer orthodontists low-cost production of virtual dental casts that can be used for digital diagnosis, treatment planning, and digital archiving of dental casts where intraoral scanners are not available.
This study aimed to measure the validity and reliability of virtual 3D orthodontic dental casts produced by smartphone-based 3D photogrammetry.
2. Material and Methods
2.1 Sample
In the study, 30 randomly selected dental casts for different orthodontic cases were selected from an orthodontic clinic. Two staff members volunteered to collect 30 orthodontic casts—15 each—from storing drawers. The casts were accessed and selected separately and by chance and were from orthodontic treatments in 2017, 2018, and 2019.
2.2 Generating photogrammetric 3D models
A Samsung Galaxy Note 9 smartphone was used to capture a set of 2D images for each cast with a zoom of × 2.0. The images were captured while the cast was rotated manually on a disk (Fig. 1A), in such a way that the images were taken around the cast in circles at different heights (Fig. 1B ). The number of images ranged from 80 to 120 within each set and were captured within a few minutes.
Fig. 1Illustration of 2D image acquisition. A dental cast was placed on a rotatable disk (A), and a set of 80 to 120 2D images with position information was captured by a smartphone (Samsung Galaxy Note 9) with zoom of × 2.0 by surrounding the cast at different heights (B).
For each dental cast, two virtual 3D images were created using two types of software: Agisoft (Metashape version 1.5.2) and 3DF Zephyr Aerial (version 4.501) (Fig. 2). First, the set of 2D images of each individual dental cast was imported into the two software programs. The process of 3D model production generally consisted of four main steps in both programs: alignment of the images, point cloud production, mesh generation, and finally, texture mapping. The constructed 3D images were exported and saved in obj.file formats.
Fig. 23D models of an orthodontic dental cast constructed using Agisoft software (left), 3DF Zephyr software (middle), CBCT (right).
The 30 casts were scanned using a cone-beam computed tomography (CBCT) device (NewTom GO), and the images from this were exported to STL.file formats. The 3D images of each dental cast from CBCT, Agisoft, and 3DF Zephyr software packages (Fig. 2) were cropped and scaled using Meshmixer 3.5.474 (Autodesk) software and saved in obj.file formats.
2.3 Measurements
To measure the accuracy of the obtained stereophotogrammetric 3D images from both programs of an individual cast, they were compared with a CBCT scan of the same cast and with each other (Fig. 3). CloudCompare software (version 2.10.2 Zephyrus) was used to compare pairings of CBCT models and stereophotogrammetric models, and also to compare each pair of stereophotogrammetric models produced by the two programs. The two 3D models of the same dental cast were imported to CloudCompare software and registered manually. Then, an iterative closest point tool was used to approximate the two surfaces of the 3D models to the minimum mean distances. A colored map of the absolute distances between the closest points of the two surfaces was created, and the distances between these points were exported to txt.file format. The means of 100%, 95th, and 90th percentiles of the absolute distances were calculated.
Fig. 3Flow chart illustrates the comparison process between virtual dental casts.
], and stereophotogrammetric images of these casts were repeated using Agioft and 3DF Zephyr software programs after a 1-month interval. The mean differences between the repeated images for each software type were calculated.
2.5 Statistical analysis
The distributions of the means of absolute distances between CBCT scans of orthodontic dental casts (the reference) and their 3D models produced by Agisoft and 3DF Zephyr software programs were tested for normality. Normality tests reject the null hypothesis that the means of absolute distances for each program are derived from a standard normal distribution (P ≤ 0.05). A Wilcoxon signed-rank test was applied to determine whether the differences between stereomodels and their CBCT counterparts and between the two programs were significantly different.
3. Results
The mean errors of repeated stereophotogrammetric images of dental casts by Agisoft and 3DF Zephyr software programs were 0.03 mm and 0.01 mm, respectively.
Table 1 shows the differences between each pair of CBCT models and the stereophotogrammetric images of Agisoft and 3DF Zephyr separately. The means of the 100% mesh point differences were 0.55 mm and 0.50 mm for the Agisoft and 3DF Zephyr software programs, respectively. The means of the 95th and 90th percentile differences were 0.44 mm and 0.38 mm for the Agisoft software, which were comparable to those of the 3DF Zephyr software (0.36 mm and 0.32 mm). All differences were not statistically significant.
Table 1Mean of absolute distances in mm and P-value of Wilcoxon signed ranks test between CBCT scans of orthodontic dental casts (the reference) and their 3D models produced by Agiosft and 3DF Zephyr software
Figure 4 shows the means of the 100%, 95th, and 90th percentiles of absolute distances between Agisoft and 3DF Zephyr programs. The differences were 0.02 mm, 0.01 mm, and 0.01 mm, respectively.
Fig. 4Box plots and whisker diagrams of the absolute distances between stereophotogrammetric models produced by Agisoft and 3DF Zephyr software. Mean, median, minimum and maximum are presented.
This study showed that 3D modeling of orthodontic dental casts using smartphone technology was applicable. The error of the constructed 3D models was about 0.3 mm in comparison with corresponding CBCT images. The mean difference between the 3D models produced by the two software types was about 0.01 mm.
The study compared the 3D models constructed by two software programs: Agisoft and 3DF Zephyr. The errors of repeated 3D models of dental casts by software programs were 0.03 mm and 0.01 mm, respectively. Accordingly, both software programs can be considered reliable.
The generated stereophotogrammetric images of dental casts produced by both software programs were compared with the corresponding CBCT models individually because the latter has been considered the gold standard [
]. The mean of the absolute distances between stereophotogrammetric images and the corresponding CBCT models was about 0.5 mm (Table 1). This difference decreased to about 0.3 mm when calculating the mean of the 95th and 90th percentiles, and the differences were not statistically significant, with errors less than the accepted clinically detectable error of 0.5 mm [
]. The decrease in the means is reasonable because the outliers had been excluded. Studies that use an iterative closest point tool for surface-based comparison usually apply the mean of the 90th percentile to eliminate the effect of outliers [
Assessment of simulated vs actual orthodontic tooth movement with a customized fixed lingual appliance using untreated posterior teeth for registration and digital superimposition: a retrospective study.
], even though the same software (Agisoft) was used in both studies. Barbero-García et al. stated that the higher errors in their study were due to the effect of hair because their images were for head vaults and not dental casts. The errors in their study could also be due to the use of a different method. In their study, the smartphone was rotated around the patients’ heads during imaging, and the 2D images were extracted from videos. In contrast, in this study, the dental cast was on a rotating disk during the imaging process, and the 2D images were directly captured. This method could give more control and reduce errors. Furthermore, the dental cast morphology was more detailed than in the patients’ head vaults; these details can help in the triangulation process and produce more accurate image registrations. The deviations from CBCT in this study were difficult to compare with the study of Stuani et al. [
] because theirs was based on the differences between linear measurements of plaster models and their digital models. They reported discrepancies in the posterior regions rather than the anterior regions, which was not the case in this study. The reason for this is not clear, but it could be due to the number of captured images for each cast of their study being less than in this study; their set of images included only 50 photos, whereas in this study, a range of 80 to 120 photos were captured for each cast. Increasing the number of captured images is crucial; an accurate and precise 3D model necessitates more informative data that improve the efficacy of the triangulation algorithm.
Studies that compared 3D images of dental casts produced by intraoral scanners with their CBCT scans showed that the differences were 0.57 mm and 0.4 mm [
. These results were comparable to the mean differences between stereophotogrammetric models and CBCT scans in our study (see Table 1). It is worth noting that these studies were landmark-based and not surface-based studies.
The color maps of this study showed that, in most of the cases, the deviation of stereophotogrammetric images from CBCT was in the interproximal areas and at occlusal details: sharp cusps, deep pits, and fissures (Fig. 5, left and middle). It has been stated that in CBCT scans, the interproximal contact points are rounded and indistinct, with a lack of detailed cuspal anatomy [
. However, the pattern of deviation between the images produced by the two software programs, in most of the cases, was in the interproximal areas. The absolute distances between the stereophotogrammetric images were less than the distances from CBCT images (Fig. 5, right). Unlike CBCT scans, the software succeeded in presenting occlusal details, but interproximal areas were difficult to reproduce with these programs.
Fig. 5Color maps show the differences between the 3D models of orthodontic dental casts. Left: the CBCT scan (the reference) and stereophotogrammetric model of Agisoft software. Middle: CBCT scan and stereophotogrammetric model of 3DF Zephyr software. Right: stereophotogrammetric models of Agisoft and 3DF Zephyr software.
The mean differences between the two software types (Agisoft and 3DF Zephyr) were minimal, at around 0.01 mm only (Fig. 4). Both programs apply the same steps and principles in 3D image production, which can reduce the errors between them. Practically, there was no difference between the two software types and both were user-friendly. The only difference was that 3D image construction was a one-step process in the 3DF Zephyr software but the steps were performed separately under the user's control in Agisoft.
Image calibration was not included in the methodology of this study because our aim was to introduce a simple technique, which is applicable to the majority rather than limited to professional dentists [
]. Unlike previous studies, dental casts used in this study were from an orthodontic clinic and were selected randomly for different orthodontic cases. The results of this study reflect the clinical application of smartphone-based 3D modeling of orthodontic dental casts for real clinical cases and are not for typodonts or dentures.
In terms of expenditure, this technique was almost free, and the requirements were feasible. The 2D images could be captured by an available smartphone device of a dental staff member, the rotating disk was inexpensive (around $15), and the 3DF Zephyr program Aerial (version 4.501) was free. In contrast, the annual cost of Agisoft software was $179, which is reasonable. Both programs have been used in the literature [
, and this study showed no significant differences between the two programs and the user desires to choose the program they prefer. Despite professional digital cameras being used in some studies [
], with a minimal cost of $200, this study proved that a smartphone camera is suitable for the generation of 3D models of orthodontic casts and can save unnecessary expenditures.
The application of this method will help to create virtual dental casts for the digital archiving of dental casts, eliminating physical storage shortcomings, such as broken models and space requirements by capturing 2D images using smartphone technology, where intraoral scanners are not available. Constantly evolving smartphone technology can open the doors to a new era in 3D imaging in dentistry and in orthodontics specifically.
5. Conclusions
The clinical application of smartphone-based photogrammetry to produce virtual orthodontic dental casts was simple, low-cost, and with errors less than the accepted clinically detectable error of 0.5 mm. Both Agisoft and 3DF Zephyr software programs were reliable for the generation of 3D models of orthodontic dental casts. Smartphone photogrammetry succeeded in presenting occlusal details, but interproximal areas were difficult to accurately reproduce.
Acknowledgment
The authors thank Dr. Abdullah Ahmed Ibrahim for helping with obtaining cone-beam computed tomography scans for the dental casts.
Author contributions
D.A. captured the 2D images of dental casts, generated the 3D models, and compared the 3D images produced by both programs. H.A. obtained the CBCT images of dental casts and measured the deviation of 3D models from CBCT images. L.Y. designed the study, performed the computations, conducted the analytic calculation, and wrote the manuscript with input from the first and second authors.
References
Crory PVM.
British Orthodontic Society's initiative on orthodontic retention, a GDP's perspective.
Assessment of simulated vs actual orthodontic tooth movement with a customized fixed lingual appliance using untreated posterior teeth for registration and digital superimposition: a retrospective study.