Table 1.

Characteristics of the studies.

Authors/Years	Type of Study	Type of AI	Materials and Methods	Results
R. Patcas et al., 2018 [32]	Observational study	ANN	Photographs of consecutive orthognathic patients were taken before and after treatment.	According to the algorithmic assessments, a significant majority of patients (66.4%) showed improvements in their appearance after treatment, resulting in an average perceived age that was nearly one year younger.
Ye-Hyun Kim et al., 2021 [33]	Observational study	ANN (ResNet-18, ResNet-34, ResNet-50 and ResNet-101)	The study included individuals who needed non-surgical orthodontic therapy and surgical orthodontic treatment.	ResNet-18 is the best model for orthognathic surgery diagnosis, providing important insights into the ideal characteristics of an AI framework for medical image-based decision-making.
Harim Kim et al., 2023 [34]	Observational study	AI-based automated assessment system	The dataset used for primary verification of the AI-based automated assessment system for Fishman’s SMI consisted of hand–wrist radiographs.	AI-based automated assessment system has proven to provide highly accurate SMI prediction with minimal errors.
Tyler Wood et al., 2023 [35]	Retrospective study	ML	Cephalometric data with Class I Angle malocclusion were utilized to train several ML methods. ANOVA was used to analyze the differences.	All of the ML systems tested properly predicted postpubertal mandibular length and Y axis of growth.
Ho Jin-Kim et al., 2022 [31]	Retrospective study	DCNN	A total of 1574 cephalometric pictures were included in the study.	The micro-average values of the DCNN-based AI model surpassed the automated tracing AI program in terms of performance.