Accuracy of artificial intelligence in the prediction of cervical vertebrae maturation stages in orthodontics: a systematic review

Objective To assess the ability of artificial intelligence in evaluating cervical vertebrae maturation stages to enhance orthodontic diagnosis considering as main outcome the accuracy of the AI software. 

Materials and methods A search was conducted of 3 databases (Cochrane Library, PubMed/MEDLINE, EMBASE) to identify studies focusing on the ability of atificial intelligence in correctly evaluating the cervical vertebrae maturation stages.  Databases were searched including articles until March 2024 only published in English. The Preferred Reporting Items for Reporting Systematic Reviews and Meta Analyses (PRISMA) protocol was adopted, two independent reviewers screened the articles and the agreement was defined by Kappa statistic. The quality of the studies was assessed through the New Castle-Ottawa scale. Due to heterogeneity of data a meta-analysis could not be performed. 

Results The search initially returned 2.953 results and after removing duplicated the number dropped to 1.104. At the end, a total of 7 studies were included in this review. It was evident that AI systems are very good in performing the screening among big amount of data, capable of differentiating what the operator often can not evaluate. 

Conclusion AI can be considered a powerful tool in helping the orthodontic diagnosis since these softwares can manage a big amount of data and perform always the same but on the other hand training of both clinicians and devices is of detrimental importance to overcome the phenomenon of overfitting and instrumental mistakes by the clinicians.