Artificial intelligence based techniques for caries risk prediction and assessment: A scoping review

Abstract

Objective

The purpose of this scoping review was to systematically search through the evidence for the applications of artificial intelligence (AI) for caries risk assessment (CRA) or prediction (CRP), determine the scope of the methodologies used, summarize their performance metrics, and report limitations and challenges (if any).

Design

A structured and comprehensive search of three electronic databases, MEDLINE, EMBASE, and Google Scholar, was performed to yield results from 2013 to 2023. Studies were selected through title, abstract, and full-text screening based on the selection criteria. Charting of the extracted data was performed using a self-designed checklist with eight dimensions.

Results

The electronic database search retrieved 3059 articles. Ultimately, 13 articles were included in the review. The most used methods were logistic regression (n = 9) and random forest (n = 8). The performance of the included models was measured variably. The reported performance metrics of the models were heterogeneous in nature; the sensitivity ranged from 0.59 to 0.996, while the specificity ranged from 0.531 to 0.943. The most frequently utilized predictors include socio-demographic factors, oral hygiene habits, and dietary habits.

Conclusion

Of the AI-based CRA models analyzed, machine learning algorithms were most frequently used. This review highlights that AI methods most probably show superior specificity and better performance than traditional methods. The application of these algorithms can have significant implications for the population impacted by pertinent chronic diseases that are avoidable through risk reduction, such as dental caries.