Deep Learning-Based Automated Diagnostic Charting on Panoramic Radiography: Comparison of YOLOv11 and YOLOv12.

Area of Science:

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

Background:

• Automated diagnostic charting on panoramic radiographs is crucial for workflow optimization and reducing diagnostic variability.
• Next-generation deep learning architectures offer potential for enhanced automated dental condition detection.

Purpose of the Study:

• To comparatively analyze the performance of YOLOv11 and YOLOv12 deep learning architectures for detecting 13 dental conditions.
• To evaluate the generalization capabilities of these models on internal and external datasets.

Main Methods:

• A hybrid dataset of 2,297 panoramic radiographs was utilized.
• Models were trained to detect conditions including caries, implants, bone loss, and impactions.
• Performance was assessed using mAP@0.5, Precision, Recall, and F1-score.

Main Results:

• YOLOv11 achieved a superior mAP@0.5 of 0.857 on the internal test set and 0.806 on the external test set.
• YOLOv11 demonstrated more robust performance and generalization than YOLOv12.
• Detection accuracy was higher for distinct objects (crowns, implants) than subtle pathologies (bone loss, caries).

Conclusions:

• YOLOv11 is a more reliable architecture for multi-class dental object detection compared to YOLOv12.
• The YOLOv11 model shows significant potential as a clinical decision support tool for enhancing diagnostic accuracy and efficiency.