CAN ARTIFICIAL INTELLIGENCE PREDICT GROWTH AND TREATMENT OUTCOMES AMONG ORTHODONTIC PATIENTS

No Thumbnail Available

Date

2021-04-16

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

The objectives of the present study were first, to synthesize the literature pertaining to artificial intelligence (AI) and machine learning (ML) applications in orthodontics, second to evaluate the possibility of predicting mandibular growth using artificial intelligence, third to assess the applicability of using artificial intelligence to predict dental treatment outcomes among Herbst patients, and finally, to predict skeletal treatment outcomes among Herbst patients using artificial intelligence. The first study was a narrative review that assessed the orthodontic literature pertaining to applications of AI and ML in orthodontics. The second study assessed the applicability of a ML method known as decision trees (DTs) for predicting maxillomandibular relationships over a five-year period using radiographs of 222 untreated subjects. The third study used DTs to predict dental treatment outcomes among 150 Herbst patients. The fourth study used a subset of 116 patients from the third study to assess possibility of using DTs to predict skeletal outcomes among Herbst patients. The first study showed that several applications of AI in orthodontics have been done, and more specifically for diagnosis and treatment planning, followed by predicting treatment outcomes, and predicting growth. The second study showed that DTs were able to successfully classify the growth of untreated subjects 85.4% of the time with the Y-axis as the most important variable for prediction. The third study demonstrated that DTs can accurately predict dental treatment outcomes among Herbst patients 81.4% of the time, and identified SN-MP, followed by overbite, and L1-MP, respectively, as the most important variables. The fourth study showed that skeletal outcomes among Herbst patients can be accurately predicted approximately 87.9% of the time. It also identified the facial convexity angle, followed by the distance from U1 to facial plane, articular angle, and Wits, respectively, as the most important variables.

Description

Keywords

Artificial Intelligence, Machine Learning, Orthodontics, Growth, Treatment Outcomes

Citation