Personalized Dental Arch Line Extraction by 3D Deep Neural Network

Abstract

Aiming at the problems of complicated interaction, low efficiency, and low degree of personalization in the extraction method of dental arch line in the existing orthodontic treatment, a personalized dental arch intelligent extraction method based on three-dimensional convolutional neural network is proposed. Firstly, the distribution of dental arches is analyzed, the sparse point cloud model is used to preprocess the dentition model, and a training data set is built. Secondly, the trained network model is used to segment the dentition point cloud, and the fully connected conditional random field is used to model and optimize the segmentation area, and the dentition point cloud with the label value of 1 in the prediction result is extracted as the preparation of the dental arc. Finally, the boundary points of the dentition point cloud with the label value of 1 in the segmentation results are used as the edge points of the preparation, and the spline curve fitting method is used to construct the shape of the dental arch line. Using 800 groups of labeled dentition point cloud training network models, the experimental results show that the segmentation accuracy of the network constructed by the proposed method can reach 96.10%, and the extraction time is shortened by 3−8 s compared with the traditional method. Compared with the manual extraction method, the average error of the dental arch line extracted by the present method is less than 0.5 mm, and the average error of the dental arch line extracted by the present method is less than 1 mm.