Deep Learning for Separation and Feature Extraction of Bonded Teeth: Tool Establishment and Application

Objective: We aimed to establish an automated digital tool for tooth surface and bracket separation and feature extraction based on deep learning algorithms and examine its application in a bracket position assessment scenario. Materials: and methods: Our segmentation network for tooth surface and bracket separation was trained using dataset A containing 978 bonded teeth, 20 original teeth, and 20 brackets generated by scanners. The accuracy and separation time of the network were tested by dataset B including another 118 bonded teeth. The clinical crown center, bracket center, and orientations of separated teeth and brackets were extracted. This tool was then applied for bracket position assessment, with the linear distribution and angular deviation of bonded brackets being recorded. Results: : This tool performed tooth surfaces and bracket separation in 2.9 ms per tooth with accuracies of 98.93% and 97.42% ( P <0.01) in datasets A and B, respectively. The features of the tooth surface and bracket were extracted and used to evaluate the results of manually bonded brackets by 49 orthodontists. Personal preferences for angulations and high consistency in locating brackets on lateral incisors were found. Conclusions: : The tool has satisfactory efficiency and accuracy and can be operated without original teeth data. It can be utilized in the bracket position assessment scenario with ease. Clinical Significance: With the help of this tool, unnecessary bracket removal can be avoided when evaluating bracket positions and changing treatment plans. It holds the promise of producing retainers and orthodontic appliances in advance without original tooth data.