Deep learning for caries detection: A systematic review.

Area of Science:

• Artificial Intelligence in Dentistry
• Machine Learning for Medical Imaging
• Dental Diagnostics

Background:

• Detecting dental caries lesions presents a significant challenge for dental practitioners.
• Deep learning (DL) models offer potential to enhance the accuracy and reliability of caries detection.

Purpose of the Study:

• To systematically review studies investigating the application of deep learning models for detecting caries lesions.
• To assess the diagnostic accuracy and quality of existing deep learning research in caries detection.

Main Methods:

• A systematic literature search was conducted across multiple databases (PubMed, Google Scholar, Scopus, Embase) and a repository (ArXiv) for studies published after 2010.
• Diagnostic accuracy studies utilizing DL models on various dental imagery were included.
• The Quality Assessment Tool for Diagnostic Accuracy Studies 2 (QUADAS-2) was employed for risk of bias assessment.

Main Results:

• Forty-two studies were included, employing classification, object detection, or segmentation DL models.
• Reported accuracies varied widely (68%-99%), with specific ranges depending on imaging modality.
• A minority of studies demonstrated low risk of bias, and 31% had low concerns regarding applicability.

Conclusions:

• The number of studies on DL for caries detection is increasing, utilizing diverse model architectures.
• While reported accuracy is promising, the overall study and reporting quality remains low.
• DL models show potential as assistive tools for identifying carious lesions.