Machine learning regression models for color prediction of CAD-CAM materials against different tooth-colored...

Area of Science:

• Dental Materials Science
• Color Science
• Machine Learning Applications in Dentistry

Background:

• Accurate color matching is crucial for aesthetic dental restorations.
• Predicting the final color of CAD-CAM materials can be challenging due to material properties, thickness, and background shade.

Purpose of the Study:

• To develop and validate a color prediction model for CAD-CAM materials.
• To assess the model's accuracy across various material thicknesses and tooth-colored backgrounds.

Main Methods:

• Four CAD-CAM materials (Lava Ultimate, Grandio Blocs, VITA Enamic, Vita Mark II) were fabricated in three thicknesses (0.5, 1.0, 1.5 mm).
• CIE-L*a*b* color coordinates were measured using a spectroradiometer against diverse backgrounds.
• Partial Least Squares (PLS) regression and leave-one-out cross-validation (LOOCV) were employed to build and test predictive models.

Main Results:

• The developed models achieved acceptable color differences (ΔE₀₀ < 1.81) for all tested conditions.
• Imperceptible color differences (ΔE₀₀ < 0.80) were obtained when data were analyzed by material and thickness.
• Predictive accuracy improved with increased material thickness, with the 'a*' coordinate showing the best model fit (RMSE = 0.14).

Conclusions:

• The integration of LOOCV and PLS regression yields a stable and reproducible predictive color model for CAD-CAM restorations.
• This model is clinically applicable for predicting the color of restorations with varying thicknesses on different tooth shades.
• Machine learning models offer a promising tool for improving the predictability and efficiency of aesthetic indirect restorations.