Machine learning in dentistry: a scoping review.

Area of Science:

• Dental informatics
• Machine learning applications
• Medical AI

Background:

• Artificial intelligence (AI), particularly machine learning (ML), is increasingly utilized in dental applications for diagnosis, prognosis, and treatment planning.
• Despite widespread adoption, a comprehensive assessment of the methodological quality and reporting completeness of these ML models in dentistry is lacking.

Purpose of the Study:

• To conduct a scoping review of published literature on ML in dentistry.
• To evaluate the methodological completeness and reporting quality of ML models using the TRIPOD + AI rubric.
• To identify key gaps and areas for improvement in ML research for dental applications.

Main Methods:

• A scoping review of PubMed-indexed articles published between January 1, 2018, and December 31, 2023, using ML in any dental specialty.
• Studies were assessed using the TRIPOD + AI checklist, focusing on data preprocessing, model validation, and clinical performance reporting.
• 1,506 articles were identified, with 280 meeting the inclusion criteria for detailed analysis.

Main Results:

• Oral and maxillofacial radiology, surgery, and general dentistry were the most represented specialties.
• A significant proportion of studies (22.9%) failed to compare their models against a clinical reference standard or existing model.
• Common limitations included insufficient bias assessment, poor outlier reporting, inadequate calibration evaluation, low reproducibility, and restricted data access.

Conclusions:

• While ML holds transformative potential for dental care, critical improvements in model calibration assessment and equity evaluation are necessary for successful clinical implementation.
• Future research should prioritize enhancing error explainability, outlier reporting, reproducibility, fairness assessments, and conducting prospective validation studies.