Research And Implementation Of Gingival Segmentation And Tooth Rendering And Display Platform For Digital Orthodontics

In recent years,with the development of 3D scanning,computer-aided design,3D printing and other technologies,digital orthodontics has developed rapidly.Because of its convenient,beautiful and comfortable wearing,it has been accepted by more and more patients,and has become the main technical means in the international orthodontic industry.Digital orthodontics requires the cooperation of professional doctors and computer-aided dental design engineers to develop orthodontic programs for patients.Therefore,a professional digital orthodontic system is necessary.At present,there are some problems in digital orthodontic system,such as tedious manual operation,low efficiency of orthodontic scheme design,poor communication between orthodontist and computer-assisted orthodontic engineer and patients.Therefore,the development of technology and system with a higher degree of automation and more convenient and intuitive communication among orthodontists,engineers and patients is of great significance for the technology improvement of our own copyright digital orthodontic system and the enhancement of international competitiveness.The process of digital orthodontics is divided into the following steps:the first step is tooth segmentation.The tooth model obtained from oral scanning is segmented,and the crown and gum as well as each tooth are divided into a separate three-dimensional model;The second step is target position prediction,which is to predict the ideal target position of the teeth according to the initial position of the segmented teeth.The third step is path planning.According to the orthodontic law,determine the number of steps in time needed to move from the initial position of the patient’s teeth to the predicted target position,and the movement path of each tooth in each step.The above three steps plan is a complete orthodontic program for patients.The fourth step is invisible braces.Firstly,the orthodontic plan of the patient is submitted to a professional dentist for review,and after confirming the feasibility of the plan,3D printing is carried out on the whole 3D tooth corresponding to each step of the tooth moving from the initial position to the target position.The 3D model of the printed tooth is used as the mold,and invisible braces are produced for the patient to wear through the press molding technology.In this paper,two key technologies of digital orthodontic system are studied:tooth segmentation and tooth rendering.Due to the personalized and diversified forms of teeth,commercial digital orthodontic software relies more on manual operation for tooth segmentation,with low efficiency.In recent years,the academic circle began to study tooth segmentation by deep learning,and made great progress in automatic segmentation by collecting a large number of patient cases.However,although the dental morphology of the collected patient cases is diverse enough,there are obvious differences between some dental models and learning samples in practical application due to the differences in dental scanning equipment and scanning methods,resulting in poor segmentation results.For the realistic rendering of tooth model,although theoretically there is a good realistic rendering model for the translucent medium,due to the need to display the tooth model to doctors and patients on the Web side,there are high requirements for quality and real-time performance,and the current commercial system can not meet the requirements technically.The main contents and contributions of this paper are as follows:(1)Studied and realized the automatic and standardized operation of dental models obtained by various dental scanning devices and scanning methods.In view of the harsh requirements of neural network-based dental model segmentation on input dental models,the standardization of dental models was innovatively proposed,and the automatic removal of redundant surfaces and positioning calibration of initial input dental models were studied.By improving the quality of input model,the precision of tooth model segmentation based on deep network is improved.(2)Studied and realized the realistic rendering of teeth based on BSDF(Bidirectional scattering distribution)model,which not only guaranteed the rendering effect but also met the requirements of real-time interaction on the Web side,providing a good look and feel and experience for dentists and patients to view the dental model.(3)Designed and developed a tooth model interactive display platform based on Web.The platform is equipped with the functions of perspective switching,dental position marking,calculation of occlusal contact points,calculation of tooth cross section,superposition of orthodontic results,deglazing of adjacent surface,viewing of dental accessories,etc.,enabling dentists to obtain dental information from various aspects and communicate with patients smoothly,thus giving professional diagnosis and treatment advice.