The primary objective of this study was to assess the accuracy and the reliability of the SureSmile OraScanner (Orametrix, Richardson, Tex) by comparing it with other desktop 3-dimensional scanners: VIVID910 (Konica Minolta, Tokyo, Japan) and R700 (3Shape, Copenhagen, Denmark). A laser-based scanner, the SLP250 Laser Probe (Laser Design, Detroit, Mich), served as the gold standard.
Five sets of dental casts were used. First, the accuracy of each scanner was studied by comparing the 3-dimensional models created by OraScanner, VIVID910, and R700 with the gold standard 3-dimensional models of the SLP250. To assess the reliability of the 3-dimensional models, the shell/shell deviation of each model was calculated based on the same surface-based registrations for all 5 sets of dental casts.
OraScanner, VIVID910, and R700 were sufficiently accurate when compared with the gold standard. In the assessment of reliability, there were no significant differences between all comparisons.
The results showed that the OraScanner system has a sophisticated algorithm for 3-dimensional surface registration and can be used to generate accurate and reliable 3-dimensional digital models for use by clinicians.
Over the past decade, the specialty of orthodontics has witnessed a marked proliferation in the use of 3-dimensional (3D) digital models for the purposes of storage, diagnosis, design of customized appliances, and orthodontic treatment outcome evaluations. Currently, a number of approaches can be applied to capture the digital representation of the 3D model. These might involve the use of structured or unstructured laser scanners to scan physical models in vitro. Also, various tabletop scanners have been designed to capture 3D images of either impressions or physical models to create 3D models. These scanners have an autorotating unit to minimize the blind area.
However, optical scanners with structured white light, such as the SureSmile OraScanner (OraMetrix, Richardson, Tex), can be used both in vivo and in vitro to scan the dentition or a physical model, respectively, to create a 3D representation. More recently, products such as iTero (Align Technology, San Jose, Calif), CEREC (Sirona, Bensheim, Germany), and Lava COS (3M Unitek, St Paul, Minn) scanners have also been introduced in the marketplace with similar functionalities. Furthermore, industrial-grade CAT scanners have been applied to scan impressions. More recently, SureSmile has extended the use of cone-beam computed tomography to capture both in-vivo images of the dentition and in-vitro images of the physical models to create 3D digital representations of the dentition. Understanding the accuracy and reliability of these imaging devices is important because they might not only affect the measures of the rendered 3D image, but also influence the accuracy and precision of any customized appliance designed on the basis of the image. Currently, there is a lack of investigations to study the accuracy and reliability of many of these scanners.
The primary objective of this study was to assess the accuracy and the reliability of the OraScanner by comparing it with other desktop 3D scanners—VIVID910 (Konica Minolta, Tokyo, Japan) and R700 (3shape, Copenhagen, Denmark)—and a laser-based scanner, the SLP250 Laser Probe (Laser Design, Detroit, Mich), which served as the gold standard. The OraScanner was studied because it has the longest history of use in orthodontics to capture both in-vivo and in-vitro 3D images of the dentition and physical models, and yet little information is available on its performance characteristics.
Material and methods
In this study, 5 sets of dental casts were used. First, all 5 sets were scanned with the OraScanner by an author (A.U.C.S.). The manufacturer of the scanner has reported that its accuracy is ±0.05 mm. The scanned data were sent automatically via the Internet with a firewall connection to the digital laboratory at OraMetrix. As part of the SureSmile service, the laboratory technicians created the digital models (shell model, gingiva model, model base, teeth model, and so on). Through the processes of denoising and refining the registration of the raw scanned data, the shell model was created; subsequently, the other models were created based on the shell model. However, only the shell models were used in this study.
Next, all 5 of the same models were scanned with the SLP250 Laser Probe. This scanning was outsourced. This was followed by desktop scanning of the same 5 models with the VIVID910 and the R700 by the same operator. The manufacturers reported that the accuracies of the SLP250, VIVID910, and R700 are ±0.01, ±0.10, and ±0.02 mm, respectively. The 3D models created by the SLP250 were used as the gold standard in this study because this scanning device has the highest accuracy compared with the others.
Cone-beam computed tomography (CBCT) was also used to assess the gold standard models as a part of this study. Ultrahigh-resolution industrial computed tomography is highly accurate. However, it is difficult to use 3D image sets with a standard personal computer because of the huge volumes of the data. So, the user-friendly CBCT data were used in this study. Trophy Pan Pro (Yoshida, Tokyo, Japan) was used; it has 0.09 mm of voxel resolution and 2.2 lp (line pairs) per millimeter of space resolution of the scan.
The 3D surface models of each dental cast were created from multiple scanned data to minimize blind sectors. Figure 1 shows the 3 scanners used in this study. The assessments of their accuracy were carried out by comparing each 3D model against the gold standard. The assessment of the reliability of the OraScanner was carried out by comparing it with each 3D model separately. This clarified the characteristics of the OraScanner by evaluating the stability of shape, distortion, and trend of magnification or reduction.
The accuracy of each scanner was studied by comparing the 3D models created by the OraScanner, VIVID910, and R700 with the gold standard 3D models created by the SLP250. The shell/shell deviation of each comparison was carried out by applying the least squares method to register each model (all 5 sets of models) using 3D reverse engineering software (Rapidform; Inus, Seoul, Korea). In this study, the maximum mean deviation set as a registration threshold was 0.01 mm that was expressed as