Evaluation of VITA shade-based tooth color categories using deep learning

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Growing patient demand for dental aesthetics necessitates objective tooth shade evaluation.
• Traditional visual shade assessment using commercial guides is subjective and lacks reliability.
• Need for advanced, objective methods in clinical dental practice.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for objective tooth shade assessment.
• To compare the accuracy of the DL model against experienced dental practitioners.
• To enhance the reliability and efficiency of dental aesthetic evaluations.

Main Methods:

• High-resolution intraoral images of 70 participants were analyzed.
• Deep learning techniques, including modified CNN architectures (ResNet), were used for tooth shade detection, segmentation, and classification.
• Performance metrics (accuracy, precision, recall, F1 score) and McNemar's test were employed for evaluation.

Main Results:

• The ResNet-based deep learning model demonstrated optimal performance in tooth shade classification.
• The DL model achieved statistically significant higher accuracy compared to experienced dentists.
• The model proved to be a reliable and efficient tool for tooth shade assessment.

Conclusions:

• Deep learning offers a reliable and objective alternative for tooth shade assessment in dentistry.
• The developed DL model minimizes subjectivity inherent in traditional visual methods.
• AI-powered tooth shade evaluation can outperform experienced clinicians, improving dental aesthetics outcomes.