Algorithms For Segmentation Of Individual Tooth Contours From Dental Computed Tomography Images

Malocclusion means misalignment of teeth between upper and lower jaws.It not only affects one's maxillofacial appearance but also affects one's maxillofacial health,causing damage to patient's body and mind.It is mainly corrected by clinical orthodontic treatment.Three-dimensional(3D)digital teeth model reconstructed from dental Computed Tomography(CT)images can provide dentists essential information to make orthodontic plan and conduct orthodontic treatment.Individual tooth contours segmentation is the key part of teeth model reconstruction from dental CT images.Due to the conglutination between adjacent teeth,teeth's being wrapped in the bone,topological difference between crown and root,segmentation of individual tooth contours from dental CT images is a technological problem of computer-aided orthodontics.We will conduct research on it.Firstly,separate adjacent teeth to extract individual tooth region from dental CT images.In order to realize that,synthesis panoramic radiography by unwrapping 3D dental CT images along dental arch.Then,detect the pixels having lower intensity to produce the separation curves between upper and lower jaws,and the separation curves between the adjacent teeth.By mapping the separation curves on panoramic radiography into 3D dental CT images to produce the separation surface between adjacent teeth,individual tooth region can be obtained.Project each individual tooth region between separation lines to two mutually perpendicular planes,and then extract the 2D axes of tooth contour on the projection planes.Obtain its 3D tooth axis by taking the 2D axes as the projection axis of the 3D tooth axis.Then,according to calculated 3D tooth axis orientation,rotate the extracted individual tooth region to improve tooth contours' similarity of adjacent CT images.Apply hybrid level set algorithm to segment tooth contours from the rotated individual tooth region.Select one necking slice from the rotated individual tooth region and take it as the initial slice.Manually segment the initial tooth contour on the initial slice,and take the segmented contour as the initial tooth contour.Then,automatically propagate tooth contour segmentation from the initial slice along two opposite directions.In the process of tooth contour propagation,tooth contour on the previous slice is taken as the prior information,and construct hybrid level set energy function to control the contour propagation.We tested our algorithm on five patients' dental CT images and results show that,our algorithm can get good results.