Table 10.

CBCT selected studies.

Authors Name and Year	Methods	Results	Authors Suggestions/Conclusions
Miki et al., (2017) [124]	AlexNet network	Accuracy = 91.0%	Automated filling of dental data for forensic identification can benefit from the suggested tooth categorization approach.
Sorkhabi and Khajeh, (2019) [125]	Customized 3D CNN	Hexagonal prism (precision = 84.63%), cylindrical voxel shapes (precision = 95.20%)	This method may help the dentists in the implant treatment from diagnosis to surgery.
Jaskari et al., (2020) [126]	Customized FCDNN	DSC were 0.57 (SD = 0.08) for the left canal and 0.58 (SD = 0.09) for the right canal	Automated DL neural network-based system when applied to CBCT scans can produce high quality segmentations of mandibular canals.
Kwak et al., (2020) [127]	2D SegNet, 2D and 3D U-Nets	2D U-Net (accuracy = 0.82), 2D SegNet (accuracy = 0.96), 3D U-Net (accuracy = 0.99)	With the help of DL, a dentist will be able to create an automated method for detecting canals, which will considerably improve the effectiveness of treatment plans and the comfort of patients.
Kim et al., (2020) [129]	CNN-based DL models	Accuracy = 93%	This method aims at assisting orthodontist to determine the best treatment path for the patient be it orthodontic or surgical treatment or a combination of both.
Orhan et al., (2020)  [130]	U-Net	Accuracy = 92.8%	AI systems based on DL methods can be useful in detecting periapical pathosis in CBCT images for clinical application.
Cui et al., (2019) [131]	Customized 3D CNN	DSC = 92.37%, DA = 99.55%, FA = 96.85%	The segmentation will fail when there is extreme gray scale value in CT image and if the tooth has the wrong orientation.
Chen et al., (2020) [132]	Multi-task 3D FCN combined with MWT	Dice = 0.936 (±0.012), Jaccard index = 0.881 (±0.019)	The multi-task 3D FCN combined with MWT can segment individual tooth of various types in dental CBCT images.
Lee et al., (2020) [133]	Fully automated CNN-based U-Net structure	Dice = 0.935, Recall = 0.956, Precision = 0.915	Some portions of the wisdom teeth were usually undetected.
Wang et al., (2021) [134]	Customized CNN	Dice similarity coefficient = 0.934 ± 0.019	DL has the potential to accurately and simultaneously segment jaw and teeth in CBCT scans.
Hiraiwa et al., (2019) [135]	AlexNet and GoogleNet	Accuracy = 86.9%	The deep learning system showed high accuracy in the differential diagnosis of a single or extra root in the distal roots of mandibular first molars.
Lee et al., (2020) [136]	GoogLeNet Inception-v3 architecture	Sensitivity = 96.1%, specificity = 77.1%, AUC = 0.91	Deep CNN architecture trained with CBCT images achieved higher diagnostic performance than that trained with panoramic images.
Ezhov et al., (2021) [137]	Customized CNN	The sensitivity values for aided and unaided groups were 0.8537 and 0.7672 while specificity was 0.9672 and 0.9616 respectively.	The proposed AI system significantly improved the diagnostic capabilities of dentists.
Qiu et al., (2021) [138]	Customized CNN	Dice (%) = 95.29	This model can be viewed as a training goal for a particular application.