Table 2.

Characteristics of the machine learning-based AI models based on intraoral and facial scanning.

Author (Year)	Application	Imaging Modality	AI Technique	Image Data Set Used to Develop the AI Model	Independent Testing Image Data Set/Validation Technique	Performance
Ghazvinian Zanjani [24] (2019)	Tooth segmentation	Intraoral scanning	CNN	120 scans, comprising 60 upper jaws and 60 lower jaws.	5-fold CV	0.94 (intersection over union score)
Kim [45] (2020)	Tooth segmentation	Intraoral scanning	Generative adversarial network	10,000 cropped images	Approximate 350 cropped images	An average improvement of 0.004 mm in the tooth segmentation
Lian [25] (2020)	Tooth labelling	Intraoral scanning	CNN	30 scans of upper jaws	5-fold CV	0.894 to 0.970 (DSC)
Liu [27] (2016)	Identification of Autism Spectrum Disorder	Facial scanning	SVM	87 scans from children with and without Autism Spectrum Disorder	LOOCV	88.51% (accuracy)
Knoops [26] (2019)	Diagnosis and planning in plastic and reconstructive surgery	Facial scanning	Machine-learning-based 3D morphable model	4261 scans from healthy subjects and orthognathic patients	LOOCV	Diagnosis 95.5% (sensitivity); 95.2% (specificity) Surgical simulation 1.1 ± 0.3 mm (accuracy)

3D, three-dimensional; AI, artificial intelligence; CV, cross-validation; DSC, dice similarity coefficient; LOOCV, leave-one-out cross-validation; SVM, support vector machine.