Artificial intelligence enabled smart phone app for real-time caries detection on bitewing radiographs

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Diagnosing proximal caries presents significant challenges in dental practice.
• Artificial intelligence (AI) is emerging as a powerful tool for medical diagnoses.
• Evaluating AI's effectiveness in dental radiography is crucial for advancing diagnostic capabilities.

Purpose of the Study:

• To assess the efficacy of an AI-powered smartphone application for real-time detection of caries lesions using bitewing radiography.
• To determine the performance metrics of the AI application in a clinical setting.

Main Methods:

• An Efficient Det-Lite1 artificial neural network was trained on 100 radiographic images.
• The trained AI model was integrated into a Google Pixel 6 smartphone app.
• The app utilized the smartphone's rear-facing camera to detect caries on 100 bitewing radiographs in real-time.

Main Results:

• The AI smartphone app demonstrated an average sensitivity of 0.75 and a precision of 0.846.
• The system successfully detected 75% of carious lesions in real-time across 100 bitewing radiographs.
• The F1 score averaged 0.795, indicating robust performance.

Conclusions:

• AI-powered smartphone applications are effective tools for real-time caries diagnosis.
• This technology offers an accessible, user-friendly, and rapid method for detecting dental caries.
• The study highlights the potential of mobile AI in improving dental radiographic interpretation.