Ortho-OPD: An Automatic Osteotomy Planes Design Model for Orthognathic Surgery Based on Deep Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence in Surgery
• Craniofacial Surgery

Background:

• Orthognathic surgery corrects dentofacial deformities, improving facial aesthetics and function.
• Virtual surgical planning (VSP) is crucial for precise, individualized orthognathic surgery.
• Manual osteotomy plane design is time-consuming and relies heavily on surgeon experience.

Purpose of the Study:

• To develop and validate an automated method for designing osteotomy planes in orthognathic surgery using deep learning.
• To improve the efficiency and precision of virtual surgical planning for craniofacial procedures.

Main Methods:

• A deep learning model, Ortho-OPD, was developed, integrating a CNN for CMF CT segmentation and the RANSAC algorithm for plane definition.
• The model was trained on 71 samples and validated on 31 cases, employing 3D boundary-sensitive loss for precision.
• Performance was assessed using metrics like Dice Similarity Coefficient (DSC) and analysis of angular and distance errors.

Main Results:

• Ortho-OPD successfully designed osteotomy planes in all tested cases with a mean DSC of 0.92 for CMF segmentation.
• The automated design process averaged approximately 9 seconds per virtual bimaxillary osteotomy, drastically reducing planning time.
• Evaluations showed no significant difference in accuracy or reliability compared to manual methods, ensuring the preservation of vital anatomical structures.

Conclusions:

• The developed Ortho-OPD system effectively automates osteotomy plane design from raw CT data for orthognathic surgery.
• This deep learning-based approach offers a highly efficient and reliable alternative to manual VSP, enhancing clinical workflow.
• Ortho-OPD represents a significant advancement in computer-assisted craniofacial surgery planning.