Photo-Based Color Analysis in Restorative Dentistry: The Role of Artificial Intelligence Algorithms

Area of Science:

• Artificial Intelligence in Dentistry
• Digital Dentistry
• Color Science in Restorative Dentistry

Background:

• Accurate tooth color assessment is crucial for esthetic restorative dentistry.
• Traditional shade selection methods can be subjective and inconsistent.
• Developing objective, digital tools for shade matching is an ongoing need.

Purpose of the Study:

• To develop and evaluate deep learning models for estimating CIELAB tooth color coordinates from extraoral photographs.
• To compare the performance of various convolutional neural network (CNN) architectures.
• To assess the clinical applicability of AI models in shade selection.

Main Methods:

• 1031 tooth images from 102 participants were used.
• CIELAB color coordinates were measured in vivo using a spectrophotometer.
• Five CNN models (CustomCNN, ResNet18, EfficientNetB0, DenseNet121, MobileNetV2) were trained and evaluated.
• Performance was assessed using Mean Absolute Error (MAE) and accuracy (ΔE ≤ 2).

Main Results:

• MobileNetV2 achieved the highest performance with >94% accuracy (ΔE ≤ 2).
• MobileNetV2, DenseNet121, and ResNet18 showed comparable, superior performance over EfficientNetB0 and CustomCNN.
• Estimation accuracy was higher for chroma and hue than for lightness and b* values.
• Anterior teeth yielded higher accuracy than posterior teeth.

Conclusions:

• AI-based models, particularly MobileNetV2, show high agreement with spectrophotometric measurements for tooth color estimation.
• These models offer potential for consistent and practical shade selection in esthetic restorative dentistry.
• AI holds promise for supporting clinical decision-making and improving consistency in dental shade matching.