ARTIFICIAL INTELLIGENCE (AI) AS AN AID IN RESTORATIVE DENTISTRY IS PROMISING, BUT STILL A WORK IN PROGRESS

Selection Criteria Besides manual search, an electronic search was conducted in 5 databases: MED-LINE, Web of Science, EMBASE, Scopus, and Cochrane. The eligibility crite-ria included clinical and in vitro studies that evaluated the diagnostic perfor-mance of artificial intelligence (AI) models in restorative dentistry to detect dental caries and vertical tooth fractures, identify tooth preparation margins, and pre-dict restoration failure. On the other hand, letters to editors, studies related to robotics in dentistry, radiographic enhancement investigations, and age estima-tion model studies were excluded. Two reviewers independently screened the title and abstract, performed data ex-traction, and assessed the risk of bias in relevant articles; the discussion with the third reviewer addressed any disagreement. The Joanna Briggs Institute JBI Crit-ical Appraisal Checklist for Quasi-Experimental Evaluation was used to appraise included studies critically for their quality. Key Study Factor A review of studies in the field of restorative dentistry that developed AI models for detecting dental caries (no = 29), vertical tooth fracture (no = 2), or the tooth finishing line (no = 1), besides studies of AI models that predict restoration failure (no = 2), were evaluated. Main Outcome Measure The diagnostic accuracy based on the sensitivity and specificity of AI models as a tool for detecting dental caries, vertical tooth fracture, the tooth finishing line, and predicting restoration failure.

Automated summary

This study comprehensively evaluated the diagnostic performance of artificial intelligence (AI) models in detecting dental caries and vertical tooth fractures, identifying tooth preparation margins, and predicting restoration failure in restorative dentistry. The eligibility criteria included clinical and in vitro studies, and a manual and electronic search was conducted in 5 databases. Two reviewers independently screened the literature, extracted data, and assessed bias, with disagreement addressed by a third reviewer.