Morphological effects of input data quantity in AI-powered dental crown design

Area of Science:

• Dental Technology
• Artificial Intelligence in Dentistry
• Digital Dentistry

Background:

• AI software is increasingly used for dental restorations.
• The quantity of input data may affect the accuracy of AI-generated dental crowns.
• Evaluating input data requirements is crucial for optimizing AI-driven dental workflows.

Purpose of the Study:

• To assess the impact of varying input data quantities on the morphological accuracy of AI-generated dental crowns.
• To determine the minimum data requirement for reliable AI-based crown reconstruction.
• To evaluate if full-jaw scans provide superior accuracy compared to quadrant or adjacent tooth data.

Main Methods:

• Retrospective in vitro study analyzing data from 30 patients.
• AI software generated crowns for a lower first molar using full dentition, quadrant, or adjacent tooth data.
• Morphological and occlusal relationships were quantitatively assessed using proposed metrics and statistical tests (Friedman Chi-Square, Wilcoxon signed rank, Kendall correlation, Fligner-Killeen).

Main Results:

• AI software reliably reconstructed dental crowns using quadrant-level data.
• Reduced input data (adjacent teeth only) led to significant morphological deviations and higher failure rates.
• No significant difference in reconstruction quality was observed between full dentition and quadrant data groups.
• Proposed metrics effectively captured distinct morphological and functional aspects of the reconstructed crowns.

Conclusions:

• Quadrant-level input data is sufficient for reliable AI-based dental crown generation.
• AI-based restoration accuracy is enhanced with adequate input data, benefiting prosthodontists by reducing chairside adjustments.
• Full-jaw scans do not offer additional accuracy benefits over quadrant data for this AI software, optimizing data acquisition.