科研成果详情

The reconstruction of the tooth 3D model is an important prior procedure in orthodontic surgery and treatment from cone-beam CT (CBCT). In this paper, we proposed a novel segmentation model for the segmentation of multiple roots tooth, which is the most challenging case in tooth segmentation, while the proposed model has a higher segmentation accuracy than conventional methods. As a whole, a slice-by-slice segmentation scheme is applied on account of the segmentation of the tooth crown with lower difficulty and the tooth roots as the hard part. In every slice, the tooth is segmented by a modified region growing method derived from the seed evolution that is maintained mainly by edge restriction and the inheritance of the segmented tooth from the previous slice. For the edge restriction, an automatic canny method is used without Gaussian blurring. From a 3D perspective, the contours of teeth in consecutive two slices are similar enough for the purpose of inheritance while some unsatisfactory cases present if the occlusion plane is not parallel to the axial plane. Therefore, we reoriented the CBCT volume with an angle generated automatically in the axial direction. To verify the accuracy of the proposed method, a total of 18 molars were used wherein the average numerical accuracy is 93.52% in term of DSC.