Developing an AI-based application for caries index detection on intraoral photographs

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Dental caries (decay) is a significant global oral health issue.
• Accurate and early detection of dental decay is crucial for effective treatment and prevention.
• Current diagnostic methods can be subjective and resource-intensive.

Purpose of the Study:

• To evaluate the effectiveness of an AI-powered smartphone application for detecting dental decay on intraoral photographs.
• To compare the diagnostic performance of the AI application against junior dentists.
• To explore the potential of AI in public health dentistry for caries assessment.

Main Methods:

• A dataset of 7,465 intraoral images was used to train a YOLOv5s model for decay detection.
• The trained model was integrated into a smartphone application.
• The AI application's performance was directly compared with that of junior dentists using 70 photographs.
• Explainable AI techniques were utilized to interpret the AI's decision-making.

Main Results:

• The YOLOv5s-based AI application achieved high performance metrics: 90.7% precision, 85.6% sensitivity, and 88.0% F1 score.
• Junior dentists achieved lower scores: 83.3% precision, 64.1% sensitivity, and 72.4% F1 score.
• The AI application demonstrated comparable or superior performance to junior dentists in decay detection.

Conclusions:

• The YOLOv5s algorithm effectively detects dental decay on intraoral photographs.
• The AI smartphone application shows potential as a reliable tool for dental professionals.
• This technology can aid in population-level caries index evaluation, contributing to reduced disease burden.