How Accurate Are Dental Models Constructed From Cone Beam Scans? Accuracy Assessment of Three-Dimensional Surface Reconstructions of Teeth From Cone Beam Computed Tomography

Background: 3-dimensional (3D) imaging, such as cone beam computed tomography (CBCT), has recently become more widely utilized. While one eventual goal of 3D imaging is to provide a means of replacing traditional dental casts, further research needs to be done in testing the accuracy of these scans when used to produce tooth models. Objective: To objectively examine the accuracy of tooth models from CBCT scans with different fields of view (FoV). Design: In vitro scans of dry human jaws. Methods: 2 dry human jaws were placed in a sealed container and submerged in water (for soft tissue simulation). Scans were performed using the Scanora 3D CBCT scanner (Soredex, Tuusula, Finland) using 3 different FoV: small (6 x 6 cm, 133 μm voxel size); medium (7.5 x 10 cm, 200 μm voxel size); and large (7.5 x 14.5 cm, 250 μm voxel size). MicroCT scans with 35 μm resolution were done on individual extracted teeth to serve as the gold standard. Individual crowns from the CBCT scans were superimposed on the MicroCT crowns to determine the error. Results: Means errors between CBCT and MicroCT were 120 μm for the small FoV, 157 for the medium FoV, and 207 μm for the large FoV. Statistically significant differences were found between the CBCT scans and the MicroCT scans as well as between the large and medium FoV scans. No significant differences were found between the medium and small FoV. CBCT models always appeared larger than the MicroCT scans. Conclusions: Error in CBCT scans can be >0.5 mm for some occlusal surfaces. While this would not be acceptable to create most prosthetic appliances, it may provide enough accuracy for diagnosis and treatment planning in orthodontics. Reviewer's Comments: Reducing or eliminating the need for impressions in the orthodontic office would be nice, especially if usable information could be extracted from a CBCT scan taken for other reasons. The method employed for this study was very time-intensive, but with sufficient accuracy, many of the steps could be automated or done by digital technicians. The relationship between error and FoV is not surprising, and it works against us in orthodontics because we would like to have a large FoV to assess skeletal structures and yet a small FoV provides the best tooth models. (Reviewer-Brent E. Larson, DDS, MS).