Development and validation of predictive models for skeletal malocclusion classification using airway and...

Area of Science:

• Orthodontics and Dental Imaging
• Artificial Intelligence in Healthcare
• 3D Medical Modeling

Background:

• Orthodontic treatment impacts airway anatomy and function.
• Cephalometric landmark analysis is crucial for skeletal classification.
• Current methods for skeletal classification may lack consistent accuracy.

Purpose of the Study:

• To develop a deep learning model for predicting skeletal malocclusions.
• To utilize airway and cephalometric landmark data from CBCT images.
• To achieve an acceptable level of accuracy in malocclusion prediction.

Main Methods:

• Retrospective collection of skeletal anatomical data in DICOM format.
• 3D model reconstruction of airway regions using 3DSlicer.
• Measurement of nasopharynx, oropharynx, and hypopharynx landmarks.
• Development of 7 artificial intelligence-based supervised learning models.

Main Results:

• The Random Forest model demonstrated the highest prediction accuracy (0.74).
• Recall scores for Class I, II, and III malocclusions were 0.71, 0.69, and 0.77, respectively.
• The model exhibited high sensitivity in correctly predicting actual positive cases.

Conclusions:

• The Random Forest model proved most accurate for predicting skeletal malocclusions.
• Airway and cephalometric landmarks are valuable inputs for AI-driven orthodontic analysis.
• Deep learning offers a viable approach to enhance diagnostic accuracy in orthodontics.