Automatic Orthodontic Treatment Planning Using Deep Learning

Intelligent dentistry is a trend in dental research and practice. In intelligent dentistry, a fundamental clinical procedure is automatic orthodontic treatment planning. Automatic orthodontic treatment planning uses algorithms to align teeth while considering oral morphology, occlusion, and biology. In clinical tooth alignment, computer-aided visualization techniques were rarely employed, instead teeth were empirically corrected in a dental cast. In the past three years, various deep learning algorithms have also been reported to make this process automatic. However, they focused on the visible morphological structure of a tooth crown. Thus, they missed critical aspects related to the full morphological framework, making their predictions clinically inconsistent for another modality. Clinically, complete planning of orthodontic treatment based on morphology ensures reliable occlusion and minimizes the biological adverse effects of apical root resorption. To fill the gap, this research develops a complete automatic orthodontic treatment planning method with three functional modules: 1) Pre-planning, including automatic tooth segmentation, identification, and model reconstruction; 2) treatment planning, including the proposed automatic tooth alignment network Deep Align Net; and 3) results for post-processing, including the stage-wise treatment plan. Finally, along with the improvements in the pre-planning stage, our approach improved the translation, rotation, and distance measurement performance evaluation metrics by 0.04 mm, 3.41°, and 0.32 mm, respectively. We achieved clinical recognition using the American board of orthodontics (ABO) protocol with an acceptable score of 6.07 out of 10. The proposed automatic orthodontic treatment planning approach included intelligent pre-treatment processing, intelligent treatment prediction and optimization, along with intelligent assistance to dental treatment implementation.