Machine Learning for Child Oral Health: A Scoping Review

Abstract

Background: Machine learning (ML) has potential to assist dental professionals with diagnosing and predicting outcomes of oral health. Tooth decay in children is the most common chronic childhood disease and it can be prevented by early detection. We aim to provide a map of the current evidence on ML in child oral health and provide insight for future research. Methods: A scoping review used databases like Medline, Web of Science, EBSCO Dentistry & Oral Science Source, Cochrane Library, and Scopus. Included studies assessed ML models for diagnoses, prediction, or management of oral health in children (0–9 years). Data extraction included publication year, location, age, sample size, disease, study type, and ML algorithms. Results: a total of 29 studies were included, mainly in pediatric dentistry and dental public health. Study designs comprised cross-sectional (34.5%), cohort (31.0%), case-control (20.7%), clinical trials (10.3%), and descriptive surveys (3.5%). The majority of studies were from high-income (69.0%) and upper middle-income countries (27.6%), with a small representation from low middle-income countries (3.4%). ML focused on predicting and diagnosing oral health issues such as caries progression and risk, with neural networks predominantly tested alongside emerging techniques like random forest, regression, and gradient boosting. Conclusions: ML algorithms hold promise in improving dental diagnosis and prediction accuracy, benefiting dental professionals, including pediatric and general dentists, in enhancing proficiency and reducing clinical errors.

Keywords

• machine learning
• child oral health
• dental diagnosis