Fast jaw segmentation from CBCT images using level-set geodesic flow with implicit shape prior

Abstract:
For the treatment planning of the patients with craniomaxillofacial deformities, cone-beam computed tomography (CBCT) is widely used and jaw segmentation can provide spatial orientation of jaw bones. On the other hand, jaw segmentation is an integral part of CBCT analysis. Therefore, it is necessary to develop a fast and accurate jaw segmentation method. In this paper, the CBCT dataset is segmented by using a thresholding method and then a novel level-set method. Our level-set formulation incorporates a modified geodesic active contour energy and a novel implicit shape prior energy. In the experiment, our proposed method was evaluated and compared with the state-of-the-art (SOTA) jaw segmentation method. These comparisons showed that our method is robust, accurate, efficient, and outperformed the SOTA method. Specifically, our average computing time for one CBCT dataset is only 3.15 seconds, 138 times faster than the SOTA method.