Automatic Point Acquisition for 3-Dimensional Cephalometric Analysis
Objectives: The increased use of 3-dimensional (3D) cone-beam computed tomography (CBCT) in orthodontic and orthognathic cases has provided a clear understanding of its benefits compared to 2-dimensional (2D) cephalometric radiographs. However, current "3D hard tissue analyses" continue to rely on 2D linear and angular measurements and do not take advantage of the vast amount of information that can be derived from CBCTs. The long-term objective is to establish a user-friendly true 3D skeletal and dental analysis including quantifiable subnorms with respect to age, gender, and ethnicity. In order to eventually create a comprehensive and effective 3D analysis, it is necessary to first develop a true 3D description of the craniofacial structure that is both accurate and practical for the user. The specific objective of this study is to develop an approach for automatic point acquisition for 3D hard tissue analysis.
Materials and Methods: The skeletal boundaries of the maxilla/palate, cranial base/zygoma, and mandible of 24 patient samples were manually plotted in a previous study conducted at UCLA. The same strict inclusion and exclusion criteria was used in both studies, including requirements for patients to be 18+ years old, have a class I skeletal relationship, and mesocephalic pattern and omitting patients with significant craniofacial defects, gross skeletal asymmetry or deformity, or history of orthognathic surgery. An extended form to the Active Shape Model (ASM) was further extended to function in 3D. Using ASM, automatic point acquisition along the boundaries of the maxilla/palate, cranial base/zygoma, and mandible was trained and implemented. A general Procrustes alignment was performed and elliptical-fourier descriptors were computed on both the points automatically plotted ("estimated landmarks") and points manually plotted ("ground-truth landmarks"), which were later compared.
Results: An approach for automatic point acquisition on normative 3D CBCT images of the adult craniofacial complex was created.
Conclusion: The development of an approach for automatic point acquisition on 3D CBCT images is an essential step towards establishing a true 3D skeletal and dental analysis. While this approach improves practicality by eliminating the need for a user to plot a great number of points thereby eliminating several hours of the user's time, further development of this method is needed to improve its accuracy.