Table 3.

Characteristics of the studies.

Authors/Years	Type of Study	Type of AI	Materials and Methods	Results
Salih Furkan Atici et al., 2023 [51]	Retrospective study	A DL network is shown, and a parallel structured DCNN with a preprocessing layer that uses X-ray pictures and age as input is proposed.	A custom CNN model with two sections, feature extraction and classification, was employed to categorize CVM into six maturation phases (CS1–CS6). AggregateNet was utilized in the model for feature extraction, while directional filters were employed as the preprocessing layer to improve the information.	AggregateNet, when combined with adjustable directional edge filters, outperformed other models with fully automated CVM stage determination.
Akay et al., 2023 [59]	Retrospective study	DL-based CNN	Digital lateral cephalometric radiographs of patients between 8 and 22 years were evaluated.	The study demonstrated that the developed model achieved moderate success.
Seo et al., 2022 [54]	Retrospective study	DeepLabv3, a semantic segmentation network for delimited cervical vertebral region, and Inception-ResNet-v2, a classification network converted to a regression model for age estimate, were used.	The study included 900 people between the ages of 4 and 18 who had a lateral cephalogram and a hand–wrist radiograph on the same day. First, the cervical vertebrae were segmented from the lateral cephalogram using DeepLabv3 architecture. Second, after isolating the region of interest from the segmented picture for preprocessing, bone age was estimated using transfer learning and an Inception-ResNet-v2 architecture-based regression model.	Using the gradient-weighted regression activation map methodology, key regions were visualized on cervical vertebral imaging to create a prediction.
Seo et al., 2021 [54]	Retrospective observational study	CNN	600 lateral cephalometric radiographs of patients aged 6–19 years; CNNs were used for CVM classification.	Achieved more than 90% accuracy in classifying CVM phases.