Three-dimensional virtual planning in mandibular advancement surgery: Soft tissue prediction based on deep learning

Purpose Artificial intelligence (AI), represented by deep learning, can be used for real-life problems and is applied across all sectors of society including medical and dental field. The purpose of this study is to review articles about deep learning that were applied to the field of oral and maxillofacial radiology. Materials and Methods A systematic review was performed using Pubmed, Scopus, and IEEE explore databases to identify articles using deep learning in English literature. The variables from 25 articles included network architecture, number of training data, evaluation result, pros and cons, study object and imaging modality. Results Convolutional Neural network (CNN) was used as a main network component. The number of published paper and training datasets tended to increase, dealing with various field of dentistry. Conclusion Dental public datasets need to be constructed and data standardization is necessary for clinical application of deep learning in dental field.

To objectively evaluate treatment outcomes in oral and maxillofacial surgery, pre- and post-treatment 3-dimensional (3D) photographs of the patient's face can be registered. For clinical use, it is of great importance that this registration process is accurate (< 1 mm). The purpose of this study was to determine the accuracy of different registration procedures. Fifteen volunteers (7 males, 8 females; mean age, 23.6 years; range, 21 to 26 years) were invited to participate in this study. Three-dimensional photographs were captured at 3 different times: baseline (T(0)), after 1 minute (T(1)), and 3 weeks later (T(2)). Furthermore, a 3D photograph of the volunteer laughing (T(L)) was acquired to investigate the effect of facial expression. Two different registration methods were used to register the photographs acquired at all different times: surface-based registration and reference-based registration. Within the surface-based registration, 2 different software packages (Maxilim [Medicim NV, Mechelen, Belgium] and 3dMD Patient [3dMD LLC, Atlanta, GA]) were used to register the 3D photographs acquired at the various times. The surface-based registration process was repeated with the preprocessed photographs. Reference-based registration (Maxilim) was performed twice by 2 observers investigating the inter- and intraobserver error. The mean registration errors are small for the 3D photographs at rest (0.39 mm for T(0)-T(1) and 0.52 mm for T(0)-T(2)). The mean registration error increased to 1.2 mm for the registration between the 3D photographs acquired at T(0) and T(L). The mean registration error for the reference-based method was 1.0 mm for T(0)-T(1), 1.1 mm for T(0)-T(2), and 1.5 mm for T(0) and T(L). The mean registration errors for the preprocessed photographs were even smaller (0.30 mm for T(0)-T(1), 0.42 mm for T(0)-T(2), and 1.2 mm for T(0) and T(L)). Furthermore, a strong correlation between the results of both software packages used for surface-based registration was found. The intra- and interobserver error for the reference-based registration method was found to be 1.2 and 1.0 mm, respectively. Surface-based registration is an accurate method to compare 3D photographs of the same individual at different times. When performing the registration procedure with the preprocessed photographs, the registration error decreases. No significant difference could be found between both software packages that were used to perform surface-based registration. Copyright © 2010 American Association of Oral and Maxillofacial Surgeons. Published by Elsevier Inc. All rights reserved.