Mandibular Angle Bone Appositions in Bruxism: A Deep Learning-Based Detection and Staging Study

Area of Science:

• Dentistry and Oral Health
• Medical Imaging and Radiology
• Artificial Intelligence in Medicine

Background:

• Bruxism, a condition characterized by teeth grinding and clenching, can lead to structural changes in the mandibular angle.
• Radiographic assessment of these bone appositions is crucial for diagnosis and management.
• Current methods may be subjective and time-consuming, highlighting the need for automated solutions.

Purpose of the Study:

• To develop and evaluate a deep learning model for automatic detection, classification, and staging of bone apposition changes in the mandibular angle region associated with bruxism.
• To assess the performance of the YOLO11x architecture in analyzing panoramic radiographs for these specific changes.
• To establish a foundation for an AI-assisted tool to support radiographic interpretation in bruxism cases.

Main Methods:

• Implementation of the YOLO11x deep learning architecture for object detection and classification.
• Annotation of 800 panoramic radiographs by a specialist into bruxism-related bone apposition stages (0-3).
• Training and validation of the model using 1600 half-jaw images, with performance metrics including mean average precision (mAP), precision, and recall.

Main Results:

• The YOLO11x model achieved high performance, with an mAP@50 of 0.864 on the validation set and 0.834 on the independent test set.
• The model demonstrated strong discriminative performance, particularly for stage 3 during training (0.901 mAP@50) and stage 0 during testing (0.909 mAP@50).
• High accuracy in anatomical localization and bounding-box precision was confirmed, with classification errors mainly occurring between adjacent stages.

Conclusions:

• Deep learning models, specifically YOLO11x, show acceptable performance in detecting and staging mandibular bone appositions related to bruxism from panoramic radiographs.
• The developed model has the potential to serve as an effective auxiliary tool for radiographic assessment and staging in clinical bruxism cases.
• Further refinement may be needed to address challenges in classifying transitional stages accurately, but the approach provides a solid basis for automated analysis.