Development of an AI-Supported Clinical Tool for Assessing Mandibular Third Molar Tooth Extraction Difficulty Using...

Area of Science:

• Oral and Maxillofacial Surgery
• Artificial Intelligence in Medicine
• Radiology

Background:

• Mandibular third molar extraction difficulty assessment is crucial for surgical planning.
• Current methods rely on subjective interpretation of radiographic features.
• An objective, AI-driven approach can improve accuracy and consistency.

Purpose of the Study:

• To develop and validate an AI-supported clinical tool for evaluating mandibular third molar extraction difficulty.
• To utilize panoramic radiographs and YOLO11 object detection models for this assessment.
• To integrate AI predictions into a user-friendly graphical interface for clinical application.

Main Methods:

• A dataset of 2000 panoramic radiographs was annotated.
• YOLO11 sub-models were trained to identify features relevant to the Pederson Index, Winter classification, and Pell and Gregory classification.
• Model performance was evaluated using precision, recall, F1 score, and mean Average Precision (mAP).

Main Results:

• YOLO11 sub-models demonstrated high accuracy across various sizes (nano to extra-large).
• Optimal sub-models were selected for Winter (angulation) and Pell and Gregory (ramus relationship and depth) criteria.
• The AI system achieved high performance metrics: 97.00% precision, 94.55% recall, and 95.76% F1 score in Pederson Index determination.

Conclusions:

• The AI tool accurately and reliably assesses mandibular third molar extraction difficulty using panoramic radiographs.
• The developed system offers a simple, effective tool for dentists to estimate extraction complexity.
• Integration into a GUI enhances clinical utility, improving decision-making and patient management.