AI-Assisted Fusion Technique for Orthodontic Diagnosis Between Cone-Beam Computed Tomography and Face Scan Data

Area of Science:

• Medical Imaging
• Computer Vision
• Machine Learning

Background:

• Accurate 3D facial reconstruction is crucial for medical imaging applications like cosmetic surgery and orthodontics.
• Integrating cone-beam computed tomography (CBCT) with facial scans presents challenges due to data acquisition differences and scan noise.
• Existing methods struggle with precise alignment and handling complex facial geometries.

Purpose of the Study:

• To develop a deep learning-based approach for accurate integration of CBCT and facial scan data.
• To enhance diagnostic accuracy and optimize treatment planning in medical imaging.
• To improve the reliability and efficiency of 3D facial recognition and alignment processes.

Main Methods:

• A deep learning model was employed for precise facial mesh detection.
• The iterative closest point (ICP) algorithm was utilized for initial alignment.
• The system was designed to handle independent data acquisition and complex facial geometries.

Main Results:

• Achieved an average registration error of approximately 0.3 mm (inlier RMSE) in feasibility testing.
• Demonstrated consistent performance across various scan orientations.
• The deep learning framework effectively processed diverse facial geometries, improving alignment reliability.

Conclusions:

• The proposed deep learning approach shows promise for accurate CBCT and facial scan integration in medical imaging.
• Preliminary results suggest potential for improved diagnostic outcomes and treatment strategies.
• Further development is needed to reduce processing time and enable simultaneous data capture for clinical validation.