Detection of mandibular fractures on panoramic radiographs using deep learning

Area of Science:

• Oral and Maxillofacial Surgery
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Mandibular fractures are common facial traumas in oral and maxillofacial surgery, impacting occlusion, function, and aesthetics.
• Accurate diagnosis and treatment are crucial for optimal patient outcomes.

Purpose of the Study:

• To develop and validate a deep learning algorithm for automated detection of mandibular fractures on panoramic radiographs (PR).

Main Methods:

• A deep learning model utilizing Faster R-CNN and Swin-Transformer was trained on 1624 manually annotated PR.
• The algorithm was validated on a dataset of 1640 PR and tested on an additional 320 PR (149 with fractures, 171 without).

Main Results:

• The deep learning model achieved a high F1 score of 0.947.
• The area-under-the-curve (AUC) for fracture detection was 0.977, indicating excellent performance.

Conclusions:

• Automated detection of mandibular fractures using deep learning shows significant promise.
• AI-assisted diagnosis can potentially reduce misdiagnosis rates and severe complications in managing mandibular fractures.