A Point Cloud Generation Network for Automatic Prediction of Postoperative Maxillofacial Soft Tissue

Area of Science:

• Biomedical Engineering
• Computer Vision
• Medical Imaging

Background:

• Accurate evaluation of postoperative soft tissue is critical for successful maxillofacial surgery.
• Restoring normal function and morphology requires precise reconstruction of soft tissue defects.

Purpose of the Study:

• To develop an advanced automatic algorithm for soft tissue defect completion.
• To enhance the accuracy and effectiveness of surgical planning in maxillofacial surgery.

Main Methods:

• A point cloud completion method utilizing Generative Adversarial Networks (GANs).
• Reconstruction of defective soft tissue areas using a dedicated network.
• Integration of generated point clouds with existing healthy tissue data.

Main Results:

• The proposed method demonstrated superior soft tissue prediction compared to existing algorithms.
• Achieved a Root Mean Squared Error of 1.45 ± 0.25 mm and Surface Distance error of 0.69 ± 0.13 mm.
• Generated soft tissue reconstructions were found to be more structurally consistent.

Conclusions:

• A novel point cloud generation network for facial soft tissue reconstruction was introduced.
• The method provides accurate and structurally consistent morphological outcomes.
• Significant potential to improve the quality and accuracy of maxillofacial surgical planning.