Machine/Deep Learning for Performing Orthodontic Diagnoses and Treatment Planning

As automated treatment planning is expected to reduce inter-planner variability and the planning time allocated for the optimization process in order to improve the plan quality, researchers have attempted to develop such systems. Artificial intelligence (AI) attempts to reflect advanced human intelligence in machines, and efforts to develop AI systems have been made since the advent of computers. In the 1980s, an expert system that expressed expert knowledge in a program produced useful results as AI technology. However, because it was a system in which rules were embedded in advance, it was limited by its difficulty in handling exceptions. Furthermore, making rules is time-consuming. The importance of machine learning that inductively learns general rules and laws based on various events (data) became recognized in the 1990s. Machine learning techniques can be said to be general-purpose techniques for detecting regularity and specificity in observational data. In this chapter, we will introduce several AI systems that have been used to derive orthodontic diagnoses and develop treatment plans in the past. We will then introduce a newly developed AI system that uses natural language processing to conduct various clinical text evaluations and develop accompanying treatment protocols.