The automatic pelvic screw corridor planning for intact pelvises based on deep learning deformable registration.

Area of Science:

• Orthopedic Surgery
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Percutaneous screw fixation in pelvic trauma is complex, relying on time-consuming, manual X-ray guided methods.
• Current screw corridor identification is labor-intensive, empirically dependent, and lacks precision.
• Surgical navigation systems offer potential simplification but require optimized planning tools.

Purpose of the Study:

• To develop and validate an automated algorithm for pelvic screw corridor planning using deep learning deformable registration.
• To enhance the efficiency and accuracy of identifying optimal screw corridors in pelvic trauma surgery.
• To improve preoperative planning and intraoperative application of screw fixation in pelvic surgery.

Main Methods:

• Proposed an automated pelvic screw corridor planning algorithm based on deep learning deformable registration.
• Incorporated corridor safety range constraints and automatic annotation of screw entrance/exit areas.
• Developed a novel, efficient algorithm for optimal corridor searching using vector-based diameter calculation.

Main Results:

• The algorithm demonstrated significant reductions in average planning time (from 1038s/3398s to 18.9s/26.7s).
• Achieved an increase in corridor diameter by 2.1%-3.3% compared to manual measurements.
• Validated on 198 intact pelvises for anterior column and S1 sacroiliac screws, showing improved efficiency and accuracy.

Conclusions:

• The deep learning-based automated planning algorithm offers a more efficient and accurate solution for pelvic screw corridor identification.
• This method has the potential to simplify complex pelvic trauma surgeries and improve patient outcomes.
• Further research is needed to validate the algorithm's performance in pelvic fracture scenarios.