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

To develop and assess the accuracy of a color prediction model for CAD-CAM materials with three thicknesses against tooth-colored backgrounds.

Specimens of Lava Ultimate, Grandio Blocs, VITA Enamic, and Vita Mark II with thicknesses of 0.5, 1.0, and 1.5 mm (n = 3) were studied. A spectroradiometer was used to obtain CIE-L*a*b* coordinates against white, black, and nine colored (ND1-ND9) backgrounds. A leave-one-out cross-validation (LOOCV) strategy was used to develop and evaluate the predictive models. Partial Least Squares (PLS) regression models were employed to predict L*, a*, and b* based on the input variables data: material, thickness, and background. Root Mean Square Error (RMSE) was used as performance assessment. CIEDE2000 color differences (ΔE₀₀) between measured and predicted data were calculated, and accuracy was assessed by comparing ΔE₀₀ values with 50:50 % acceptability and perceptibility thresholds.

The integration of the predictive methods resulted in acceptable (ΔE₀₀<1.81) color difference values for all models and imperceptible ΔE₀₀ (ΔE₀₀<0.80) when data were separated by material and thickness. The best model fit was consistently observed for the a* coordinate, showing the RMSE values closer to zero, with the separation of input data on material and thickness demonstrating the highest predictive adjustment due to the lowest RMSE=0.14. Predictive models performed better for greater thicknesses.

The developed integration of LOOCV and PLS regression produced a stable, reproducible, and clinically applicable predictive model for CAD-CAM restoratives with varying thicknesses and across different tooth-colored backgrounds, demonstrating acceptable color differences.

Integrating the machine learning regression models into clinical workflows may be a promising tool for achieving more predictable, efficient, and high-quality aesthetic indirect restorations, even when dealing with varying abutment tooth shades.