Detection of Jaw Lesions on Panoramic Radiographs Using Deep Learning Method

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Jaw lesions, both radiolucent and radiopaque, require accurate detection on panoramic radiographs.
• Deep learning models offer potential for improving diagnostic accuracy in dental radiology.

Purpose of the Study:

• To evaluate and compare state-of-the-art deep learning models for detecting and segmenting jaw lesions.
• To assess model performance under different training scenarios, including spatial localization.

Main Methods:

• Retrospective collection of 2371 anonymized panoramic radiographs with jaw lesions.
• Training and evaluation of YOLOv8, YOLOv11, Mask R-CNN, and RT-DETR models.
• Performance assessment using precision, recall, F1-score, and mean average precision (mAP).

Main Results:

• YOLOv11x-seg and YOLOv8x-seg showed high segmentation performance in Scenario I.
• RT-DETR-L demonstrated superior detection performance, especially in Scenario III with spatial localization.
• Deep learning models showed potential for effective lesion detection and segmentation.

Conclusions:

• Deep learning models can effectively detect and segment jaw lesions on panoramic radiographs.
• RT-DETR-L shows strong potential, particularly when trained with spatial localization data.
• Models are recommended as clinical decision support tools, not standalone diagnostic systems.