Maxillary crowding and spacing: validation of an artificial intelligence model vs. digitally assisted human observer

Area of Science:

• Dentistry
• Orthodontics
• Artificial Intelligence

Background:

• Quantifying dental arch crowding and spacing is crucial for orthodontic diagnosis.
• Manual analysis of tooth size-arch length discrepancy (TSALD) can be time-consuming and subjective.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for quantifying upper arch crowding and spacing.
• To compare the AI model's accuracy against manual digital space analysis.

Main Methods:

• An AI model (YOLO 8 Pose) was trained on 832 intraoral images and occlusal scans.
• The AI model performed automated tooth segmentation, point detection, and TSALD quantification.
• AI-derived TSALD measurements were compared with manual analysis using OrthoCAD software on 300 test images.

Main Results:

• The AI model achieved 90% overall accuracy in quantifying TSALD.
• Strong correlation (r > 0.92) was observed between AI and manual TSALD measurements.
• Discrepancies were noted in mild crowding (7%) and moderate to severe spacing (3.3-5%).

Conclusions:

• A validated AI model can reliably quantify TSALD in orthodontic diagnostics.
• The AI model demonstrates potential interchangeability with manual analysis methods.
• This technology can enhance efficiency and accuracy in orthodontic treatment planning.