Artificial Intelligence-Based Multi-Stage System for Automated Angle's Classification of Malocclusion from Intraoral...

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate Angle's classification of occlusion is crucial in orthodontics.
• Manual classification is time-consuming and prone to variability.

Purpose of the Study:

• To develop and validate a multi-stage deep learning pipeline for automated Angle's classification of occlusion using intraoral images.
• To assess the pipeline's accuracy, speed, and clinical applicability.

Main Methods:

• A pipeline integrating Occlusion Side Classification (OSC), side-specific MolarBBoxNet, and AngleClassifier-R50 was developed.
• Trained on 8909 intraoral occlusion images, validated on 383 unseen images.
• Grad-CAM visualizations were used for interpretability.

Main Results:

• Perfect occlusion side classification (accuracy 1.00) and high molar classification accuracy (97.41% internal, 94.3% external).
• High accuracy for Class I (99.0% sensitivity, 98.9% specificity), Class II (88.2% sensitivity, 98.7% specificity), and Class III (97.2% sensitivity, 93.4% specificity).
• Processing time of ~0.11s per image, significantly faster than manual annotation.

Conclusions:

• The deep learning pipeline accurately and efficiently classifies Angle's occlusion from intraoral images.
• The system demonstrates potential for general practice, tele-orthodontics, large-scale screening, and reducing diagnostic variability.
• Future work may involve multimodal inputs for comprehensive orthodontic diagnostics.