Article Preview
TopIntroduction
In the digital information age, with the application of all kinds of electronic sensors, the data that people can obtain is becoming more and more fine, and the medical industry has always been an industry closely related to people's life, and a large number of medical data are produced all the time (Fiori et al., 2016; Conrad et al., 2002; Pestian et al., 2006), among which oral medical data are deeply concerned by patients, doctors and researchers. On the one hand, this kind of data is real-time, and doctors need to deal with it in time to develop a reasonable and efficient treatment plan for patients (Tan et al., 2021); on the other hand, such data are closely related to patients' privacy, and hospitals or research institutions need to establish intelligent management methods to strictly restrain the dissemination, sharing, and use of such data (Tan et al., 2022; Wang et al., 2021). Therefore, how to use and manage these data efficiently and safely has become a hot topic (Yang et al., 2021).
One of the uses of oral medical data is dental correction. Traditional orthodontic methods generally use plaster models to simulate the teeth and jaws of patients, but this plaster model has many inconveniences in preservation and use, such as breakage and breakage. It takes up space, which is not conducive for doctors to understand the dental conditions of patients in different treatment stages, and finally affect the effect of dental correction. At the same time, the orthodontic appliance used in the traditional correction method also has the shortcomings of long correction time, complex wearing mode and easy to cause. These factors lead doctors and patients to hope for new ways of correction.
In recent years, with the development of computer graphics technology and the improvement of modern medical level, computer aided design system has been widely used in more and more fields (Yu et al., 2021; Rodby et al., 2014; Tian et al., 2021; Reighard et al., 2021; Dovramadjiev et al., 2021; Rakishev et al., 2022). Virtual orthodontic system plays an important role in the field of oral medicine. The system uses the data of dental model collected by 3D scanning equipment as input to record and save the shape and parameters of teeth. A computer was used to simulate the entire course of the patient’s orthodontic treatment, recording the changes in tooth position at each stage. Based on these data, the invisible braces were designed at different stages, Patients simply need to wear different invisible braces at different times according to the design, can make the upper or lower occlusal teeth move according to the design, Eventually, the patient’s teeth become even. Patients can also see the virtual treatment process before treatment and know the correction results in advance.
It is an important preprocessing of virtual orthodontic system to accurately segment teeth from dental models (Kau et al., 2011; Petrescu et al., 2022; Choi et al., 2021). Its purpose is to separate the single crowns from the dental models, and separate the crowns from each other. It lays the foundation for the movement and arrangement of teeth. At present, many 3D model segmentation methods have been proposed in the field of computer graphics, but there are still some challenges in the complete automatic segmentation of 3D models. Although teeth have obvious geometric features, teeth of different people have obvious morphological differences, and due to the influence of scanning accuracy and 3D mesh reconstruction accuracy, there are often uneven and adhesive areas on the surface of dental models. These factors make it difficult for traditional methods to achieve completely automatic and accurate segmentation.
According to the minimum rule of human vision theory, the dividing line between teeth and gums is located in the region with large negative curvature of the model. The methods based on curvature information usually detect these negative curvature regions and segment teeth. However, due to the uneven areas between teeth and gums, these areas will cause serious interference to the segmentation method based on curvature information, and it is difficult to achieve accurate segmentation of the model.