Automated Cephalometric Points Marking System

Area of Science:

• Medical image analysis
• Artificial intelligence in dentistry
• Orthodontic diagnostics

Background:

• Accurate cephalometric landmark detection is crucial for orthodontic diagnosis and treatment planning.
• Automating this process using artificial intelligence (AI) presents significant challenges in medical and dental imaging.
• Existing methods require improvement for clinical relevance.

Purpose of the Study:

• To develop and evaluate an enhanced AI-based approach for automatic cephalometric landmark detection.
• To improve the accuracy and reliability of landmark identification in orthodontic X-ray images.
• To support clinical decision-making in orthodontics through advanced image analysis.

Main Methods:

• A multi-model strategy was developed, integrating an ALD algorithm with three derived models.
• Extensive image augmentation techniques, including contrast and negative transformations, were employed for model training.
• An ensemble approach combined outputs from all models, selecting the best prediction based on performance.

Main Results:

• The proposed multi-model approach achieved a mean radial error (MRE) of 2.12 mm, outperforming the baseline model (2.26 mm).
• A successful detection rate (SDR) of 72.22% within a 2.5 mm threshold was achieved, exceeding the baseline model's 68.87%.
• The ensemble method demonstrated superior performance in cephalometric landmark detection.

Conclusions:

• The ensemble-based approach significantly enhances the accuracy of cephalometric landmark detection.
• This AI-driven method shows strong potential for integration into clinical orthodontic workflows.
• Improved landmark detection accuracy can lead to more precise orthodontic diagnoses and treatment plans.