Research On Automatic Tooth Segmentation In Dental CBCT Images

Dental cone beam computed tomography(CBCT),with the characteristics of clear imaging and low radiation dose,has become an indispensable imaging technology in the research of dental diseases and dental problems.The accurate and automatic segmentation of individual tooth is critical for computer-aided analysis towards clinical decision support and treatment planning.Three-dimensional reconstruction of individual tooth after the segmentation also plays an important role in the simulation in digital orthodontics.According to the structures of individual tooth,tooth is composed of enamel,dentine,and dental pulp.The segmentation and understanding of the morphological structures of each tooth component play an important role in the attention of dental health and root canal treatment.However,several difficulties are encountered in the segmentation of individual tooth in dental CBCT images,which are due to the similar intensities between tooth and alveolar bone,and the blurring boundaries of neighboring teeth.The small volume of enamel,dentine,and dental pulp,as well as the intra-and inter-intensity variabilities form the challenges in segmenting tooth structures.We aim to solve these segmentation difficulties and realize the segmentation of individual tooth as well as the segmentation of tooth structures.The main research work of this paper includes:(1)Establishment of the dental CBCT image database.There is no publicly available database of dental CBCT images,and the tooth segmentation results require comparative evaluation with tooth ground truth.In this paper,25 cases of dental CBCT images are manual delineated to generate the tooth ground truth for conducting the research of individual tooth segmentation.Then,the model of individual tooth segmentation is performed to the remaining 24 dental CBCT images for segmenting individual tooth.With a small amount of manual delineations for revising the segmentation results,we obtain the tooth ground truth.The method of combining the segmentation model with a small amount of delineations greatly improves the efficiency of tooth ground truth generation,and effectively expands the tooth database.(2)Automatic segmentation of individual tooth from dental CBCT images based on a 3D multi-task fully convolutional network(FCN).The multi-task FCN learns to simultaneously predict the probability of tooth region and the probability of tooth surface.The foreground marker and the background marker are obtained from the tooth probability map.By combining the tooth probability gradient map and the surface probability map as the input image,marker-controlled watershed transform(MWT)is used to automatically separate and segment individual tooth.Twenty-five dental CBCT scans were used in the study.The average Dice similarity coefficient,Jaccard index,and relative volume difference are 0.936(±0.012),0.881(±0.019),and 0.072(±0.027),respectively,and the average symmetric surface distance is 0.363(±0.145)mm for our method.The experimental results demonstrate that the multi-task 3D FCN that is combined with MWT could automatically segment individual tooth of various types in dental CBCT images.The method can also segment teeth in a non-open bite position and handle the dental CBCT image with metal artifacts(3)Automatic segmentation of tooth structures based on Ostu algorithm.Based on the individual tooth segmentation results,we use the Ostu algorithm as a core algorithm to segment tooth structures.In this work,the algorithm is firstly performed to segment the enamel.After removing the enamel from the individual tooth,Ostu algorithm is conducted to the remaining structures for segmenting the dentine and dental pulp.The method achieves promising segmentation results.