Detection of Dental Anomalies in Digital Panoramic Images Using YOLO: A Next Generation Approach Based on Single...

Area of Science:

• Artificial Intelligence in Dentistry
• Medical Imaging Analysis
• Deep Learning for Diagnostic Support

Background:

• Diagnosing pediatric dental conditions from panoramic radiographs is challenging due to mixed dentition, leading to inconsistent interpretations.
• Developing objective tools is crucial for improving diagnostic accuracy and efficiency in pediatric dentistry.

Purpose of the Study:

• To develop and validate an advanced deep learning model for enhanced diagnostic accuracy in pediatric dental panoramic radiographs.
• To identify the optimal YOLO variant for detecting common pediatric dental findings.

Main Methods:

• Comparative analysis of YOLOv8, v9, v10, and v11 architectures for detecting Dental Caries, Deciduous Tooth, Root Canal Treatment, and Pulpotomy.
• Two-tiered validation using a primary dataset (n=644) for training and an independent external dataset (n=150) for testing.
• Annotation and validation by a dual-expert team (pediatric dentist and oral/maxillofacial radiologist).

Main Results:

• YOLOv11x was selected as the optimal model, achieving a mean Average Precision (mAP50) of 0.91 on the internal validation set.
• On the external test set, YOLOv11x demonstrated robust generalization with an overall F1-Score of 0.81 and mAP50 of 0.82.
• High recall rates were achieved for Root Canal Treatment (88%), Pulpotomy (86%), Deciduous Tooth (84%), and Dental Caries (79%).

Conclusions:

• The YOLOv11x model is a validated, accurate, and reliable tool for automated detection of pediatric dental findings in panoramic radiographs.
• AI-driven systems like YOLOv11x can serve as valuable assistive tools, supporting clinical decision-making and diagnostic workflows.
• This technology contributes to the consistent detection of common dental conditions in pediatric patients.