Performance of Artificial Intelligence in diagnosing dental caries- An umbrella review

Abstract

Introduction: Nearly 3.5 million people worldwide suffer from dental caries. Untreated dental caries has been identified as a prevalent oral health condition that raises the burden on oral health and may cause other systemic illnesses. Smartphones and Artificial Intelligence are now widely used in many healthcare domains, including diagnosis and treatment planning. So, this umbrella review aims to appraise the existing evidence on the effectiveness of various algorithms of Artificial Intelligence in diagnosing dental caries from previously published systematic reviews.

Material and Materials: A broad electronic search was done among various databases that focus on existing systematic reviews on dental caries detection or diagnosis using various algorithms of Artificial Intelligence. The quality assessment of the included systematic reviews was done using the AMSTAR-2 checklist and the risk of bias was done using the ROBIS tool.

Results: Among 503 reviews, 9 were included in this umbrella review. Based on the AMSTAR-2 checklist, 7 reviews were stated as high quality and 2 reviews as moderate quality; and the all the reviews were stated to have low risk of bias. The sensitivity, and specificity values of machine learning models employed in diagnosing dental caries ranged from 71% to 99% and 90% to 94% respectively, and accuracy ranged from 77.4% to 90%, whereas the sensitivity, specificity values for deep learning models ranged from 78.6% to 87% and 78.4%. to 87.8% respectively.

Conclusion: The diagnostic performance of artificial intelligence models varies between machine learning and deep learning algorithms. Based on these findings, machine learning has shown promising results in diagnosing dental caries