During cone beam computed tomography (CBCT) scanning, intra-oral metallic objects may produce streak artefacts, which impair the occlusal surface of the teeth. This study aimed to determine the accuracy of replacement of the CBCT dentition with a more accurate dentition and to determine the clinical feasibility of the method. Impressions of the teeth of six cadaveric skulls with unrestored dentitions were taken and acrylic base plates constructed incorporating radiopaque registration markers. Each appliance was fitted to the skull and a CBCT performed. Impressions were taken of the dentition with the devices in situ and dental models were produced. These were CBCT-scanned and the images of the skulls and models imported into computer-aided design/computer-aided manufacturing (CAD/CAM) software and aligned on the registration markers. The occlusal surfaces of each dentition were then replaced with the occlusal image of the corresponding model. The absolute mean distance between the registration markers in the skulls and the dental models was 0.09 ± 0.02 mm, and for the dentition was 0.24 ± 0.09 mm. When the method was applied to patients, the distance between markers was 0.12 ± 0.04 mm for the maxilla and 0.16 ± 0.02 mm for the mandible. It is possible to replace the inaccurate dentition on a CBCT scan using this method and to create a composite skull which is clinically acceptable.
It is widely acknowledged that the use of computed tomography (CT) in the oral and maxillofacial area has introduced a new epoch in dentistry. CT allows clinicians to access the internal hard tissue structures of the head and neck region in a virtual environment. When displayed in 3-dimensional (3D) format, valuable spatial information about the patient in a variety of directions, as well as cross-sections, can be obtained. Cone beam computed tomography (CBCT) is a relatively recent innovation and has very similar characteristics to conventional CT but with less radiation exposure to the patient. CBCT was specifically designed for use in the maxillofacial region for the visualization of hard tissue.
As with conventional CT, CBCT does not record the occlusal surfaces of the teeth accurately. When imaging patients using CBCT, any intra-oral metallic objects (e.g. restorations, jewellery, implants, and orthodontic appliances) create streak artefacts. This occurs because the density of metal is outside the normal range that can be processed by the CBCT software, resulting in incomplete attenuation profiles. These artefacts can obliterate the occlusal surfaces on the images of the teeth, rendering the virtual model inaccurate in predicting intercuspal relationship and in constructing occlusal guiding wafers for orthognathic surgery. Several methods have been developed in an attempt to remove these artefacts, ranging from metal artefact reduction (MAR) algorithms requiring specialized software, which need to be clinically validated, to extra-oral registration techniques that distort the soft tissue around the area of the mouth and lips during scanning. None of the craniofacial imaging techniques currently available can simultaneously capture facial soft tissues, the facial skeleton, and dentition at an optimal quality for clinical use. This may be achieved by imaging each structure independently, using various imaging modalities, and then fusing the images into a single composite image.
In the present study, an image fusion method that replaces the inaccurate occlusal surfaces of the teeth from a CBCT-generated image with a more accurate occlusal surface, without any distortion of the surrounding soft tissues, was described and evaluated.
The first aim of this study was to determine the accuracy of replacement of the virtual dentition in dry human maxillae and mandibles with corresponding dentitions derived from scanned plaster study casts using a custom-made intra-oral reference device. The second aim was to assess the feasibility of applying the same technique clinically to replace the distorted dentition recorded by the CBCT using the intra-oral transfer device.
Materials and methods
Construction of a transfer device
Upper and lower impressions of the dentition of six dried human skulls with complete unrestored dentitions were taken using alginate (Alginot™, Kerr Corporation, Romulus, MI, USA). These were cast in a high density minimal expansion gypsum product (Shera Hard Rock, SHERA Werkstoff-Technologie, Lemförde, Germany). The dental models were coated with a separating medium (sodium alginate) and a self-curing polymethyl methacrylate (John Winter & Co. Ltd, Halifax, UK) base plate constructed on the palatal and lingual aspects of each of the upper and lower models. Three hexagonal radiolucent markers were constructed using a class IV minimal expansion gypsum product (Hard Rock) and secured to each base plate ( Fig. 1 ) producing an intra-oral transfer device (IOTD).
The six dried mandibles were laser-scanned and CBCT-scanned with and without the IOTD in situ using a NextEngine desktop 3D scanner and ScanStudio software (NextEngine, Santa Monica, CA, USA) ( Fig. 2 ). According to the manufacturer’s documentation, the scanner is accurate to 0.005 mm. This was regarded as the gold standard for capturing a 3D image of the mandibles. Prior to each mandible being scanned, the system was calibrated using the automated calibration process. Each image was saved as mesh surface data (STL format). The CBCT scan was performed using an iCAT scanner (Imaging Sciences International, Hatfield, PA, USA), at a setting of 20 s, 0.4 mm voxel size.
Alginate impressions of the dentitions were taken with the IOTD still in situ and plaster models were produced; these were also laser-scanned and CBCT-scanned, but at a setting of 20 s, 0.2 mm voxel size.
The CBCT volumetric data of the skull hard tissue and the plaster models (DICOM format) was converted into STL format using MeVisLab (MeVis Medical Solutions Ltd., Bremen, Germany). The software utilizes the marching cubes algorithm to extract the polygonal mesh of an isosurface from the voxel data.
Determination of the accuracy of the CBCT image of the jaw bones and dentition
The rationale behind this part of the study was to quantify magnification errors associated with CBCT scanning of the jaw bones; the laser-scanned image was used as the gold standard to quantify this potential source of errors. The mesh surface models were imported into VRMesh (Seattle City, WA, USA).
To determine if the proposed technique was accurate, the CBCT-scanned bone image of the mandible was superimposed over the laser-scanned bone image of the same mandible, having first virtually ‘hidden’ the dentition from both images, so they were not involved in the alignment stage of the bone. The superimposition was initiated using rigid registration based on landmarks common to both images, and the alignment was further refined using surface-based registration based on the ICP (iterative closest point) alignment. The absolute mean distance for 90% of the points making up the bone surface meshes was recorded. Following bone alignment using rigid registration, the dentition was made visible and the absolute mean distance for 90% of the points making up the dentition mesh surface was recorded ( Fig. 3 ).
Determination of the accuracy of the CBCT image of the plaster dentition
This determined the magnitude of error of the CBCT scanning of plaster models using the laser scanned image as the gold standard. The STL file for each plaster model captured by laser scanning and CBCT were imported into VRMesh. Corresponding landmarks were identified on each image and manual rigid alignment was initially applied followed by fine alignment. The absolute mean distance for 90% of the points between the surfaces was recorded.
Replacement of the dentition
Each CBCT mandibular image (0.4 mm resolution) and corresponding plaster model image (0.2 mm resolution), both with the IOTD in situ , were imported into VRMesh. The hexagonal markers on the IOTD were identified and manually aligned using VRMesh; the superimposition of the markers was then refined using ICP. Since the markers were common to both images, surface alignment of the markers only would align the dentition without relying on the teeth themselves for superimposition. This allowed the occlusal surface of the dentition of the mandible to be replaced with the occlusal surface image from the plaster models. To assess the accuracy of the technique, the absolute mean distance for 90% of the points between the two meshes was recorded for the hexagonal markers and also between the two dentitions, i.e. the mandibular dentition of the jaw scan and the corresponding dentition of the model scan. The same procedure was repeated on the six maxillary dentitions and corresponding plaster dentitions.
In order to assess the practicality of using the IOTD in clinical practice, ethical approval was obtained and six patients requiring surgical correction of their dentofacial deformities were recruited. All patients had undergone pre-surgical orthodontic treatment using conventional fixed appliances that were still present. As part of their routine pre-operative work-up, each patient was fitted with both upper and lower IOTDs; these were checked for fit and retention, and once secure, a 22 cm, 0.4 mm voxel CBCT scan was taken. Following the scan, upper and lower alginate impressions were taken with the IOTDs still in situ and plaster models produced. The same methodology of dentition replacement was applied to the 12 dentitions and corresponding plaster dentitions ( Fig. 4 ).