Predicting the Debonding of CAD/CAM Composite Resin Crowns with AI

Area of Science:

• Dental Materials Science
• Artificial Intelligence in Dentistry
• Biomedical Imaging

Background:

• Debonding of computer-aided design/computer-aided manufacturing (CAD/CAM) composite resin (CR) crowns reduces their survival rate.
• Currently, no established preventive measures exist to mitigate this issue.
• Accurate prediction of debonding is crucial for improving clinical outcomes.

Purpose of the Study:

• To evaluate the efficacy of a deep learning model employing a convolution neural network (CNN) for predicting the debonding probability of CAD/CAM CR crowns.
• To assess the model's performance using 2D images derived from 3D stereolithography models of dental dies.
• To establish a novel AI-driven approach for preventing CAD/CAM CR crown debonding.

Main Methods:

• A dataset of 24 CAD/CAM CR crown cases (12 debonded, 12 trouble-free) was utilized.
• 8,640 images from 3D stereolithography models were randomly allocated into training/validation (6,480) and test (2,160) sets.
• A deep learning CNN model was developed and trained to predict debonding probability.

Main Results:

• The CNN model achieved high prediction accuracy (98.5%), precision (97.0%), recall (100%), and F-measure (0.985).
• The model demonstrated excellent diagnostic capability with an area under the curve (AUC) of 0.998.
• The mean calculation time was remarkably fast at 2 ms/step for 2,160 test images.

Conclusions:

• Deep learning with CNNs shows significant potential for accurately predicting CAD/CAM CR crown debonding.
• This AI-based method offers a viable strategy to enhance the longevity and survival rates of dental restorations.
• The study highlights the successful application of AI in dental prosthodontics for predictive diagnostics.