The Application of Machine Learning to Analyze the Modalities of Surgery-First and Orthodontic-First Approach in Orthognathic Surgery

The concepts of surgery-first approach (SFA) in orthognathic surgery (OgS) was applied based on the regional accelerated phenomenon, the orthodontic tooth movement could be rapid during the first 3-4 months after surgery. The lengthy conventional concepts of dental decompensation can be achieved by 2-jaw rotational surgery. According to our previous researches to compare the SFA with conventional approach in class III patients by 3-facial dimension, the surgical stability, esthetic, occlusion and final outcome. There were not different between two approaches. The construction of 3D craniofacial model and computer-aided surgical simulation provide precision and predictability of OgS. However, the principle of dental occlusion setup was mainly based on preference and experience of the orthodontists who setup the bite in an orthodontic treatable malocclusion. The procedure to transfer the occlusion into craniofacial models requires manual setup of dental models and 2-3 times of scanning procedures, causing problems of errors of stone model fabrication, infection control and material waste. The ultimate goal should be totally digital, replace dental model with intraoral scan and automatic digital surgical occlusion setup. Neural network machine learning applies mathematic computation on a big data to optimize the learning method and strategies. On the other hand, the scope of human sense recognition in artificial intelligence, could be applied to make human decision through mathematic method. It is a powerful tool to solve clinical problems such as classification of patterns and morphology, diagnosis, treatment decision, cephalometric landmark identification and process of 3D images. The aim of the study was to: (1) to recognize the occlusal characteristics between the cases with SFA and orthodontic first approach (OFA); (2) to automatically classify of the treatment modes with machine learning. It is a 2-year project. The samples consisted of 200 cases of surgical-orthodontics with 3D surgical simulation in the data base. The samples are classified into 2 groups; SFA and ODA for 100 cases in each group. The data collection included demographic data, general dental health, history related to facial trauma and craniofacial anomalies, panoramic radiographs (before and after OgS), STL file of 3D dental models with presurgical bite registration and surgical setup, color mapping of the bite registration. The dental characteristics will be measured by Ortho AnalyzerTM . Meanwhile, machine learning will be applied to analyze the data and automatic case classification. In the first year, digital model preparation and standardization of data for machine learning will be accomplished; in second year, data analysis and model generation will be conducted.

Project ID:PC10907-1116
External Project ID:MOST109-2314-B182-012