Artificial intelligence system for automatic deciduous tooth detection and numbering in panoramic radiographs.

Area of Science:

• Artificial Intelligence in Dentistry
• Pediatric Dental Radiology
• Deep Learning Applications

Background:

• Accurate dental charting is crucial for diagnosis and identification.
• Manual tooth numbering on panoramic radiographs can be time-consuming and prone to error.
• Automated methods are needed to improve efficiency and accuracy in pediatric dental assessments.

Purpose of the Study:

• To evaluate a deep-learning algorithm for automated detection and numbering of deciduous teeth.
• To assess the performance of an AI system on pediatric panoramic radiographs.

Main Methods:

• Development of an AI algorithm using Faster R-CNN Inception v2 (COCO) models.
• Training and testing the algorithm on 421 pediatric panoramic images.
• Performance evaluation using a confusion matrix to determine sensitivity, precision, and F1 score.

Main Results:

• The AI system demonstrated high accuracy in detecting and numbering deciduous teeth.
• Achieved a sensitivity of 0.9804, precision of 0.9571, and F1 score of 0.9686.
• The deep-learning approach proved effective for automated dental charting.

Conclusions:

• Deep-learning AI models offer a promising solution for automated dental radiographic charting in children.
• AI tools can save clinicians time and improve diagnostic aid.
• Automated dental charting has significant potential in forensic identification applications.