ARTIFICIAL INTELLIGENCE MODELS SHOW POTENTIAL IN RECOGNIZING THE DENTAL IMPLANT TYPE, PREDICTING IMPLANT SUCCESS, AND OPTIMIZING IMPLANT DESIGN

Selection Criteria Up until February 21, 2021, an electronic search was undertaken in 5 databases: MEDLINE, Web of Science, EMBASE, Scopus , Cochrane. A manual search was also carried out. Included were clinical , in vitro studies that evaluated the performance of Artificial intelligence (AI) models in implant dentistry for im-plant type recognition, implant success prediction using patient risk factors and ontology criteria, and implant design optimization using finite element analysis (FEA) calculations and AI models. However, investigations of non-dental implant-related AI applications and review papers were excluded.Key Study Factor Review of studies that developed AI models for implant type recognition, implant success prediction, and implant design optimization. The included models are based on Classical Machine Learning (eg Decision tree learning, Random forest, Logistic regression, and Multidimensional unfolding analysis) or Artificial Neural Networks models such as Deep Convolutional Neural Networks and Residual Neural Networks. Seventeen articles were included in the qualitative analysis. These articles were sorted into different AI applications, namely AI models for implant type recognition (no =7), prediction models for implant success forecast (no =7), and models for optimization of implant designs (no =3). Main Outcome Measure AI model diagnostic accuracy for recognition of the implant type, forecast of the implant success by using patient risk factors and ontology criteria, and optimiza-tion of implant designs by combining FEA calculations and AI models.Main Results Seventeen articles were included in the qualitative analysis. The AI models created to detect implant type using periapi-cal , panoramic radiographs achieved an overall accuracy of 93.8%-98%. The models used to predict osteointegration success or implant success differed across research, ranging from 62.4% to 80.5%. The research that created AI models to optimize implant designs seems to agree on the useful-ness of AI models to enhance dental implant design, such as reducing stress at the implant-bone interface by 36.6% when compared to the FEA model, optimizing implant de-sign porosity, length , diameter, improving FEA calcula-tions, or accurately determining the elastic modulus of the implant-bone interface. Nine of included investigations were considered as a low risk of bias and eight as high risk accord-ing to Joanna Briggs Institute JBI Critical Appraisal Checklist for Quasi-Experimental Evaluation.Conclusions The authors conclude that the performance of AI models for implant type detection, implant success prediction, and im-plant design optimization has shown significant promise but is still in progress, and they encourage more study in this field.