Table 6.

Advanced AI applications in periodontology.

Authors	Summarized Abstract	Methods Used	Results	Conclusions
Chau [51]	-AI may be useful for providing automated visual plaque control advice based on intraoral photographs.	-Collection of intraoral photographs labeled as healthy, diseased, or questionable. -Analysis of accuracy in gingivitis detection using artificial intelligence system.	-AI system had sensitivity of 0.92 and specificity of 0.94. -Mean intersection over union of the system was 0.60.	-Potential for monitoring patients’ plaque control effectiveness using AI system.
Lin [52]	-Proposes automatic alveolar bone loss measurement system for periodontitis diagnosis.	-TSLS and ABLifBm for teeth contours and bone loss areas. -CEJ_LG method for CEJ, ALC, and APEX localization.	-More than half of bone loss measurements within 10% deviation -All bone loss measurements within 25% deviation from ground truth.	-The proposed system effectively estimates horizontal alveolar bone loss. -The system can aid in early and accurate diagnosis of bone loss.
Chang [53]	-Develops hybrid method for staging periodontitis using deep learning architecture.	-Hybrid framework of deep learning and conventional CAD processing. -Automatic detection and classification of periodontal bone loss.	-Mean absolute differences between periodontitis stages were insignificant. -Overall ICC value between developed method and radiologists’ diagnoses was 0.91.	-Hybrid framework shows high accuracy in automatic periodontitis diagnosis. -Automatic method has high reliability compared to radiologists’ diagnoses.
Thanathornwong [54]	-Deep learning detects periodontally compromised teeth in digital panoramic radiographs.	-Utilized a faster R-CNN for periodontally compromised teeth detection. -Model used a pretrained ResNet architecture for detection.	-Faster R-CNN detected periodontally compromised teeth with 0.81 precision. -Model excluded healthy teeth areas, showing a recall rate of 0.80.	-Application of a faster R-CNN reduces diagnostic effort and enables automated screening.
Lee [55]	-Develops deep CNN algorithm for diagnosing and predicting periodontally compromised teeth.	-Combined pretrained deep CNN architecture with self-trained network. -Used periapical radiographic images for optimal CNN algorithm.	-Deep CNN algorithm had 81.0% diagnostic accuracy for premolars. -Predicted extraction accuracy was 82.8% for premolars.	-Deep CNN algorithm useful for diagnosing and predicting periodontally compromised teeth.
Alotaibi [56]	-Develops CNN algorithm for bone loss detection in dental radiographs.	-Utilized pre-trained deep CNN architecture and self-trained network. -CNN (VGG-16) used for detecting alveolar bone loss in radiographs.	-Deep CNN algorithm detected alveolar bone loss with 73% accuracy. -Model classified severity of bone loss with 59% accuracy.	-Deep CNN algorithm useful in detecting alveolar bone loss. -Machines can perform better in image diagnosis with deep learning.