Dental pulp segmentation from cone-beam computed tomography images

When dental pulp needs to be removed in individuals who require root canal therapy, dental pulp segmentation from cone-beam computed tomography (CBCT) images plays a vital role in assisting clinical decisions by making through simulation of dental pulp removal. Dental pulp has complex topological shapes and inhomogeneous intensity distribution, so we propose a level set method to incorporate the piecewise area energy with the local edge energy into semiautomatically segment dental pulps in the three-dimensional domain. The minimization of the piecewise area energy approximates a separation between the target and the intricate background region. In the local edge energy, our edge indicator function highlights the dental pulp boundaries with weak ramps. We compared our approach with the start-of-the-art method of dental pulp segmentation, the well-known DRLSE model for image segmentation, and the manual delineation by the trained operator. The DRLSE model was not successful in segmenting dental pulps of some teeth, and our method outperformed the start-of-the-art method. Four quantitative metrics were applied between our method and the manual delineation, and our results on 29 dental pulps showed that our approach has a root-mean-square surface distance of 1.22 ± 0.31 mm, 3.13 ± 1.35 mm, 2.63 ± 1.92 mm, 2.03 ± 0.82 mm, and 2.06 ± 1.48 mm in dental pulps of wisdom teeth, molars, premolars, canines, and incisors, respectively.