Automatic Segmentation of Bone Graft in Maxillary Sinus via Distance Constrained Network Guided by Prior Anatomical...

Area of Science:

• Biomedical Imaging
• Oral and Maxillofacial Surgery
• Artificial Intelligence in Medicine

Background:

• Maxillary sinus lifting is vital for dental implants, requiring precise analysis of bone graft (BG) geometry.
• Automated segmentation of BGs in cone beam computed tomography (CBCT) is challenging due to image complexities.
• Existing tools lack efficiency and accuracy for BG segmentation.

Purpose of the Study:

• To develop an automated method for accurate bone graft segmentation in the maxillary sinus.
• To enhance quantitative analysis of geometric changes in bone grafts for improved dental implant outcomes.

Main Methods:

• A deep neural network (DNN) incorporating preoperative anatomical knowledge for guidance.
• A coordinate attention gate to enhance feature identification in skip connections.
• Geodesic distance constraints integrated into the DNN for multi-task prediction.

Main Results:

• Achieved a Dice similarity coefficient of 85.48 ± 6.38%.
• Demonstrated an average surface distance error of 0.57 ± 0.34 mm and a 95% Hausdorff distance of 2.64 ± 2.09 mm.
• Outperformed comparison networks in segmentation accuracy and efficiency.

Conclusions:

• The proposed DNN effectively segments bone grafts in the maxillary sinus.
• This method offers potential for precise analysis of bone graft volume changes and absorption rates.
• Improved segmentation accuracy supports better clinical decision-making in dental implantology.