AUTOMATIC SEGMENTATION OF THE MANDIBLE FROM LIMITED-ANGLE DENTAL X-RAY TOMOGRAPHY RECONSTRUCTIONS

A 3-D reconstruction from sparse limited-angle X-ray projection data is a useful compromise between a single radiograph and a full CT reconstruction, e.g. in dental imaging. The segmentation of such volumes is desirable for clinical applications such as implantology, but the task is complicated by the inherent limited spatial validity of the reconstructions. We present an automatic model-based method for extracting the mandible from 3-D limited-angle dental X-ray reconstructions. The process includes enhancing the reconstruction, estimating the successfully reconstructed mandibular area, and the actual segmentation process. The results with 13 reconstructions are good with an average segmentation error of 0.32 mm