Introduction
Three-dimensional (3D) virtual treatment planning is becoming an increasingly important tool within the fields of oral and maxillofacial surgery, orthodontics, and implantology [
1‐
3]. Complex anatomical structures and the relations between these structures can be visualized using cone-beam computed tomography (CBCT). More specifically, bony structures and soft tissues can be captured using CBCT, which can be used to create a virtual head model [
4,
5]. In addition, 3D stereophotogrammetry can be used to add texture and detail to the virtual head model [
6]. To acquire accurate information about the dentition, additional imaging of the occlusal surfaces, e.g., intra-oral scans or plaster casts, is needed because of scattering present in a CBCT due to the high density of enamel, dental restorations, implants, and orthodontic appliances [
7‐
9]. Accurately capturing the dentition is of major importance as drilling guides, saw guides, or orthognathic positioning guides are often dental occlusal-surface-supported [
10‐
12]. Apart from accuracy, the dentition should also be positioned at the correct anatomical position in the mandible and maxilla. Therefore, accurate matching of the dentition in the CBCT is required for the use in clinical practice.
Several methods are described in literature to solve the issue of a distorted occlusal area within the CBCT scan. Swennen et al. [
13] proposed a triple CBCT scan method using voxel-based registration to capture accurate occlusal surfaces. However, this requires the patient to be scanned twice, increasing the radiation exposure. Other methods have investigated the image-fusion of digital dental models with the CBCT using fiducial superimposition [
14‐
17] or surface-based fusion [
18,
19]. The study performed by Swennen et al. [
13] is, to the best knowledge of the authors, the only study that utilized commercially available software.
The use of different imaging modalities (CBCT and plaster cast/intra-oral scans) allows surgeons and technicians to virtually plan and practice several treatments before the actual treatment takes place. Three-dimensional virtual orthognathic surgical planning is a widely used tool to plan and simulate different treatment options [
3]. Additionally, in dental implantology, digital implant planning is also widely used to preoperatively assess the bone quality and digitally plan the optimal implant position. Three-dimensional printed surgical drilling guides can be used to transfer the digital plan towards the operating theater [
20]. Within the orthodontic work field, 3D techniques are used for several different applications. Orthodontic virtual setups can be created and used to assess the accuracy of the treatment. [
21] Digitally created indirect bonding trays can be used for optimal bracket placement and an enhancement of the workflow [
22].
In the current workflow, taking physical impressions, and pouring them into plaster casts, is the most used technique to capture the maxillary and mandibular occlusal surfaces precisely [
23‐
28]. With the introduction and technical evolution of intra-oral scanners, it became easier and quicker to obtain a detailed model of the dentition of the patient without the need for physical impressions. Earlier studies proved that an intra-oral scanner is a valid method for accurately visualizing the dentition. Furthermore, patients generally experience an intra-oral scan as a more comfortable way of getting an accurate 3D dentition compared to physical impressions [
28‐
31].
The purpose of this study is to assess the accuracy of the replacement of distorted dentition in CBCT scans with accurate digital dental models, using commercially available software.
Results
A total of ten dry skulls were used in the study. All skulls had ≥ 24 teeth in order to achieve a proper fusion of the intra-oral scan in the CBCT. Data distribution was tested for normality using the Shapiro-Wilk test and showed a normal distribution (p > 0.05).
Gold standard
Aligning the structured light scan (gold standard) towards the CBCT using the ICP algorithm in the 3DMedX software, using the forehead, orbital rims and zygomatic area as reference, resulted in an average matching accuracy of 0.20 ± 0.16 mm for the skull/maxilla and 0.11 ± 0.12 mm for the mandible.
Accuracy
The translational directions investigated were left/right, anterior/posterior, and cranial/caudal. For the rotational directions, the Pitch (rotation around the
x-axis), Roll (rotation around
y-axis), and Yaw (rotation around
z-axis) rotations were used. Translational and rotational directions are displayed in Table
2 for both IPS CaseDesigner® and OrthoAnalyzer™. Corresponding
p values are also given in Table
1 as well as the
p values for the statistical differences between IPS CaseDesigner® and OrthoAnalyzer™.
Table 2
Accuracy of the fusion of the intra-oral scan with the CBCT compared to the gold standard expressed in the translational directions x (left-right), y (anterior-posterior), z (cranial-caudal), overall accuracy and rotational directions Pitch (rotation around x-axis), Roll (rotation around y-axis), and Yaw (rotation around z-axis)
IPS Case Designer® | Maxilla | 0.29 ± 0.19 0.01 | 0.25 ± 0.31 0.01 | 0.40 ± 0.28 0.02 | 0.60 | 0.25 ± 0.20 0.01 | 0.29 ± 0.28 < 0.01 | 0.18 ± 0.13 < 0.01 |
| Mandible | 0.30 ± 0.31 0.01 | 0.36 ± 0.40 0.01 | 0.45 ± 0.46 < 0.01 | 0.66 | 0.32 ± 0.24 < 0.01 | 0.22 ± 0.20 0.01 | 0.16 ± 0.12 < 0.01 |
Ortho-Analyzer™ | Maxilla | 0.47 ± 0.29 0.00 | 0.62 ± 0.53 0.00 | 0.83 ± 0.54 0.00 | 1.18 | 0.26 ± 0.20 0.00 | 0.45 ± 0.35 0.00 | 0.28 ± 0.29 0.01 |
Mandible | 0.41 ± 0.44 < 0.01 | 0.62 ± 0.49 < 0.01 | 0.78 ± 0.76 0.01 | 1.09 | 0.37 ± 0.47 0.02 | 0.21 ± 0.21 0.01 | 0.23 ± 0.22 < 0.01 |
p value** | Maxilla | 0.65 | 0.06 | 0.26 | – | 0.71 | 0.04 | 0.34 |
Mandible | 0.01 | 0.08 | 0.18 | – | 0.85 | 0.57 | 0.13 |
| 0.19 | 0.37 | 0.43 | – | 0.01 | 0.16 | 0.10 |
| 0.11 | 0.28 | 0.33 | – | 0.05 | 0.01 | 0.07 |
For IPS CaseDesigner®, translational differences compared to the gold standard were all smaller than the voxel size of the CBCT (0.40 mm) except for the cranial/caudal direction which was the largest in the mandible (0.45 ± 0.46 mm). Rotational differences compared to the gold standard did not exceed 0.32° with the yaw of mandible showing the smallest difference (0.16 ± 0.12°) and the roll of the maxilla the largest (0.29 ± 0.28°). The paired t test showed a statistically significant difference between the gold standard and IPS CaseDesigner® for all translations and rotations. However, the differences were ≤ 0.40 mm except for the cranial-caudal translation.
For the OrthoAnalyzer™ software, the translational differences were the largest for cranial/caudal direction of the maxilla (0.83 ± 0.54 mm) whereas the left/right direction of the mandible showed the smallest inaccuracy (0.41 ± 0.44 mm). Rotational differences did not exceed 0.45°, with the roll of the mandible showing the smallest difference (0.21 ± 0.21°) and the roll of the maxilla the largest (0.45 ± 0.35°). Translational and rotational differences all statistically differ from the gold standard for the OrthoAnalyzer™.
IPS CaseDesigner® showed the smallest overall difference for the maxilla (0.60 mm) whereas the maxilla of the OrthoAnalyzer™ showed the largest difference (1.18 mm).
Comparing both software packages, OrthoAnalyzer™ showed bigger translational discrepancies compared to IPS CaseDesigner®. Only the left/right direction of the mandible (p = 0.01) showed a statistically significant difference. Rotational differences were all smaller than 0.50° for all ten skulls, and only the Roll of the maxilla showed a statistically significant difference (p = 0.04).
Reproducibility
The intra-observer overall mean difference was 0.08 ± 0.10 mm for IPS CaseDesigner® and 0.18 ± 0.15 mm for OrthoAnalyzer™. The inter-observer overall mean difference was 0.11 ± 0.12 mm for IPS CaseDesigner® and 0.21 ± 0.19 mm for OrthoAnalyzer™. The intra- and inter-observer reliability analysis showed good correlation and intraclass coefficients for OrthoAnalyzer™. High correlation and intraclass coefficients were found for IPS CaseDesigner®. In comparison with IPS CaseDesigner®, OrthoAnalyzer™ showed lower scores. The results of the analysis are shown in Table
3.
Table 3
The intraclass correlation coefficient displayed for all translations and rotations for both the maxilla and mandible
IPS Case Designer® | Maxilla | Intra-observer ICC | 0.97 | 0.98 | 0.97 | 0.93 | 0.95 | 0.97 |
Inter-observer ICC | 0.94 | 0.98 | 0.92 | 0.92 | 0.97 | 0.90 |
Mandible | Intra-observer ICC | 0.99 | 0.99 | 0.98 | 0.99 | 0.98 | 0.97 |
Inter-observer ICC | 0.97 | 0.97 | 0.93 | 0.99 | 0.99 | 0.96 |
Ortho Analyzer™ | Maxilla | Intra-observer ICC | 0.90 | 0.84 | 0.89 | 0.89 | 0.83 | 0.92 |
Inter-observer ICC | 0.84 | 0.85 | 0.82 | 0.88 | 0.88 | 0.90 |
Mandible | Intra-observer ICC | 0.89 | 089 | 0.92 | 0.90 | 0.92 | 0.93 |
Inter-observer ICC | 0.88 | 0.84 | 0.89 | 0.90 | 0.90 | 0.93 |
Discussion
CBCT imaging is a widely used tool for capturing the human skull. However, CBCT has the drawback that it is prone to distortions around the dentition. Metallic restorations, orthodontic appliances, and the high density of enamel cause distortion of the dentition in the CBCT model [
7‐
9]. In order to utilize CBCT imaging for CAD/CAM processes, additional imaging is needed as well as a proper fusion between the different imaging modalities. In a recent review of Mangano et al. [
34], it was concluded that there was still no easy way to fuse scans from different image modalities.
In this study, the accuracy of the fusion of intra-oral scans into CBCT models was assessed using commercially available software. The tested software packages, IPS CaseDesigner® and OrthoAnalyzer™, showed a high level of accuracy compared to the gold standard. The accuracy was calculated for all six degrees of freedom. It was noticeable that the accuracy in the cranial/caudal direction was the lowest for IPS CaseDesigner® and OrthoAnalyzer™ in both the maxilla and mandible. A logical reason for this lower accuracy could not be found, but it is worth to note this difference. The user of these software packages should take this larger inaccuracy in the cranial/caudal direction into account when performing a surgical planning. A visual check whether the software performed an accurate fusion is strongly advisable.
An important step in assessing the fusion accuracy is the alignment of the structured light scan of the skull (gold standard) with the CBCT 3D model. This was performed utilizing a validated ICP algorithm [
32]. The accuracy of the alignment of the gold standard with the CBCT 3D model was 0.20 mm. This is a clinically acceptable result as the resolution of the CBCT scans was 0.40 mm.
Overall mean intra- and inter-observer differences were low (≤ 0.21 mm) which was reflected in the ICC values found. In the OrthoAnalyzer™, the user needs to provide three corresponding points on both the CBCT 3D model and the intra-oral scan. As the results show, this increases the intra- or inter-observer variability in OrthoAnalyzer™ compared to the workflow in IPS CaseDesigner which did not use additional manual input. However, all ICC values were > 0.82 showing good agreement for OrthoAnalyzer™ and > 0.92 for IPS CaseDesigner® showing excellent intra- and inter-observer agreement.
When comparing the fusion techniques of this current study to earlier studies, it is noteworthy that all earlier studies either utilize intra-oral markers [
17,
35,
36], extra-oral markers [
15,
37], or utilized a double/triple scanning procedure [
13]. The fusion technique utilized by IPS CaseDesigner® and OrthoAnalyzer™ does not require markers or an additional CBCT scan. This makes it a convenient method for use in the clinical practice. Moreover, the accuracy found for IPS CaseDesigner® is in line with other studies. A splint with ceramic balls was used by Uechi et al. [
15], and a root-mean-square error of 0.4 mm was found. Another study found an accuracy ranging from 0.10 to 0.50 mm by using titanium markers [
37]. De Waard et al. found errors ranging from 0.12 to 0.45 mm. Another study performed by Lin et al. using surface-based matching found errors ranging from 0.11 to 0.53 mm. However, most of these studies assessed the accuracy of the fusion (e.g., how do the markers overlap) instead of assessing the differences using a true gold standard. The accuracies found for IPS CaseDesigner are in the range of these studies as the overall accuracy is ≤ 0.66 mm. OrthoAnalyzer shows bigger discrepancies (≤ 1.18 mm) and is therefore not in line with earlier studies.
A limitation of the current study is the use of dry human skulls which could influence the result of the study. For example, accuracy of intra-oral scanning can be lower in actual patients as patient movement, anatomical restrictions, and excessive saliva can hamper proper imaging of the dentition [
28]. Another limitation of the current study was the absence of orthodontic appliances. Most orthognathic patients have orthodontic brackets which can influence the accuracy of the fusion as might cause distortion of the CBCT scan. Therefore, a future study to investigate the effects of orthodontic appliances is necessary. Furthermore, to study the use of the fusion of intra-oral scans in CBCT in implantology patients closer, a study should be designed in which partial dental arches are used to see whether the fusion is still accurate if more teeth are missing.
Recent developments in artificial intelligence (AI) are promising. Earlier studies showed that using AI, it is possible to segment third molars from an orthopantomogram [
38]. With future developments, AI might be a promising technique to automatically “recognize” dentition in a CBCT image. Recognizing the dentition might make it easier to replace it and therefore enhancing the accuracy of the fusion between intra-oral scans and CBCT. As AI (e.g., convolutional neural networks) is an upcoming and promising technique in image fusion [
39], development of AI-driven algorithms to fuse dental information with CBCT data might result in a more accurate and automated solution.
Conclusion
IPS CaseDesigner® is a reliable software packages within the scope of this study. It provides accurate fusion of the intra-oral scan in the CBCT when a complete dental arch is used with little to none dental fillings. OrthoAnalyzer™ showed bigger discrepancies, and therefore, it is recommended to perform proper visual inspection before using the fusion. Inaccuracies were found in both packages. However, for IPS CaseDesigner, these are in line with the findings of similar studies. OrthoAnalyzer shows bigger discrepancies. Future research towards the effect of scattering caused by fillings and orthodontic appliances is recommended as well as the influence of missing teeth.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.