Detecting the presence of taurodont teeth on panoramic radiographs using a deep learning-based convolutional neural...

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Taurodontism is a developmental anomaly of teeth requiring accurate diagnosis.
• Analyzing panoramic radiography for taurodontism can be challenging for clinicians.
• Artificial intelligence (AI) offers potential for improved medical image analysis.

Purpose of the Study:

• To develop and evaluate a CNN-based AI model for diagnosing taurodontism in panoramic radiography.
• To assess the performance of the AI model in segmenting taurodont teeth.

Main Methods:

• A U-Net model implemented in Pytorch was used to develop automatic taurodont tooth segmentation.
• 434 panoramic radiography images from individuals over 13 years old were utilized.
• Data augmentation techniques and a Confusion Matrix were employed to train and evaluate the model.

Main Results:

• The AI model achieved a sensitivity of 0.8650, precision of 0.7898, and F1-score of 0.8257 for taurodont tooth segmentation.
• The model demonstrated high accuracy in identifying taurodontism, comparable to expert performance.

Conclusions:

• CNN-based AI models show significant promise for diagnosing taurodontism in panoramic radiography.
• The developed AI system can assist clinicians in accurate taurodontism detection and potentially therapeutic decision-making.