Reconstruction of Three-dimensional Craniofacial Image with High-Accuracy Dentition Images

Many patients with congenital malformation or jaw deformity exhibit craniofacial deformity. Since precise examination is essential for these cases, three-dimensional (3D) simulation models of craniofacial structure reconstructed from multi-detector computed tomography (MDCT) are frequently used. However, oral orthodontic appliances and the occlusion of upper and lower teeth make reproduction of the exact tooth shape difficult. Thus, trials are underway to generate 3D models in which the dentition reconstructed from digital imaging and communications in medicine (DICOM) data (CT data) is replaced by laser-scanned images. However, the integration accuracy and the optimum conditions for integration of MDCT and laser-scanned images are not known. Examining these points, a 3D model of craniofacial structure with accurate dentition was reconstructed in this study.<br>As materials, test pieces having CT values comparable to those of tooth, bone tissue and soft tissue were prepared. CT data and stereolithography (STL) data obtained from a 3D non-contact laser scanner were compared with data from an industrial μCT (true values). The optimum thresholds of test pieces were determined by comparison with those of true values. The integration accuracy was examined both by point and surface-based registration.<br>The different thresholds showed a wide variation in images of test pieces reconstructed from CT data, indicating that the optimum threshold is an important factor for accuracy. Using each optimum threshold, test pieces of tooth, bone tissue and soft tissue showed a high accuracy with a difference from the true value of less than 113.2μm. The difference between images from STL data and the true value was also less than 42.8μm. The surface-based registration was more reliable than the point-based registration at superimposition. In images utilizing the optimum threshold and the surface-based registration, the superimposition error between CT and STL data of the dry mandible was 173.8 and 257.8μm at the dentition and at the bone surface, respectively. The error of the patient's craniofacial structure was less than 233.8μm in the whole dentition. The present study demonstrates that acceptable craniofacial 3D images with tooth contact can be generated by optimizing various conditions.