Positional assessment of lower third molar and mandibular canal using explainable artificial intelligence

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Assessing the positional relationship between lower third molars (M3i) and the mandibular canal (MC) is crucial for surgical planning and preventing nerve damage.
• Panoramic radiographs (PRs) are commonly used, but manual assessment can be subjective and time-consuming.

Purpose of the Study:

• To develop and validate an automated system using artificial intelligence (AI) for assessing the positional relationship between M3i and MC on PRs.
• To classify the relationship into three distinct categories for improved clinical decision-making.

Main Methods:

• A deep learning model (MobileNet-V2) combined with skeletonization and signed distance methods was trained on 1444 manually annotated M3s from 863 PRs.
• The model was validated on a separate dataset of 130 PRs (217 M3s), and performance metrics including accuracy, precision, and F1-score were calculated.

Main Results:

• The AI system demonstrated high performance with a weighted accuracy of 0.951.
• Key metrics included precision (0.943), sensitivity (0.941), specificity (0.800), negative predictive value (0.865), and F1-score (0.938).

Conclusions:

• AI-enhanced assessment of PRs provides an objective, accurate, and reproducible method for determining the M3i-MC positional relationship.
• This explainable AI system can assist clinicians in intuitive positional assessments and warrants further research for predicting alveolar nerve injury risk.